Part III - Racing Tactics and Strategy
Build 27 · 20 September 2026 · not yet reviewed
12Racecourse Strategy
Sections in this chapter
Strategy is the plan for sailing the course as if no other boats were present; tactics are the execution of that plan among them S51. This chapter treats the plan as a hypothesis: a statement about where the wind, current and waves will favour the boat, held and attached to the observations that would prove it wrong. Section 12.1 defines the hypothesis and its . Sections 12.2 to 12.5 supply its inputs: the wind field, the current field, the wave field and the geometry of the course and its management. Section 12.6 ranks those inputs when they disagree, section 12.7 describes how the plan is updated under way without losing decisiveness, and section 12.8 closes the loop by comparing the pre-race prediction with what the racecourse actually did, so that each race sharpens the next model. Throughout, the crew's own logged values are the reference wherever the published record is silent, and every figure quoted carries its source.
12.1Pre-race racecourse model
The racecourse model is the crew's single, ranked answer to which way the first beat pays, by what mechanism and with what . It is produced ashore from forecast, tide and venue knowledge, then revised on the water from direct observation before the warning signal. Its purpose is to make the start plan and the first-beat plan follow from a cause rather than a hunch, and to name in advance the evidence that would overturn it. A model that cannot be overturned by any observation is not a model but a preference.
12.1.1Defining the initial strategic hypothesis. The hypothesis has four parts: the expected mechanism (oscillating instability, a persistent rotation, a current gradient, a shoreline bend), the side or mode it favours, the confidence the crew attaches to it, and the time window over which it holds. Walker's framework, as reviewed by SailZing, sorts every leg into three categories, an oscillating wind, one side obviously advantaged, or everything else, where the boat sails conservatively toward the middle and consolidates S62. Assigning the leg to one category is the first act of the hypothesis, because the decision rules differ: tack on headers in the first, commit toward the favoured side in the second, and protect position in the third.
12.1.2Identifying the most important uncertainty before launching. Every model rests on one assumption that carries more of the result than the others, and the crew names it before leaving the shore. On a sea-breeze day the assumption is whether the breeze arrives before the start: Houghton's rule expects one when an offshore is under 25 knots with thin or broken cloud S43, and the breeze needs a land-sea temperature difference of 5 °C or more S42. On a tidal venue the assumption is the time of the turn relative to the sequence. The observation plan on the water is aimed at that assumption first, because resolving it changes the plan; resolving the others merely refines it.
12.1.3Recording wind direction, pressure distribution, current, waves, and geometry. The record is made in the boat's own units so it can be compared later: compass headings on each tack and the mean between them, the range of directions seen, the position and movement of darker water, the set and drift read from fixed references, the wave direction relative to the wind, and the bearing of the start line. permits timing devices and up to two compasses with heading, heading memory and timing functions only S3, so the record is written or memorised in the format the crew has agreed. Every entry carries a time, because a heading without a time cannot distinguish a shift from a trend.
12.1.4Evidence that confirms or invalidates the model. Each mechanism produces a signature the crew can check. An oscillating hypothesis predicts headings that swing around a stable mean with a period, Bethwaite's measurements giving about 3 minutes in heavy air and 10 minutes or more in light-medium air S50; a mean that walks steadily one way over three or four cycles invalidates it and points to a persistent mechanism. A sea-breeze hypothesis predicts the breeze filling first offshore and moving toward land as a S51; a fill from the shore side invalidates it. A current hypothesis predicts a consistent lean on moored objects; a change in that lean before the predicted turn invalidates the tide table for that spot.
12.1.5Translating observations into start and first-beat implications. The model fixes the side; the start position follows the side. Dellenbaugh's rule places the boat at the pin for a left-side plan and at the committee boat for a right-side plan, and lets a strongly favoured side override the line bias S51. The line bias is then a modifier, not the decision: for every 5 degrees of upwind bias to one end, that end is closer to the windward mark by about one eighth of the line length in distance sailed S65. The first-beat implication is the order of the first three moves: which tack to hold off the line, which shift or pressure feature triggers the first tack, and which boundary of the course the crew will not cross.
12.1.6Gold-medal standard: arrive with a prioritised model, not unfiltered observations. Observed from the coach boat, the crew reaches the starting area able to state, in one sentence each, the favoured side, the mechanism, the confidence, the one uncertainty being watched and the observation that would flip the plan. Asked at the preparatory signal, the helm gives the same answer the crew gave at the warning signal S1, or a different answer with the new observation that caused it. No observation is reported that has not been , and no plan element exists without a named .
12.2Wind-field structure
The wind the boat sails in is the modified by the surface and the shoreline, then modulated by instability and by clouds. ECMWF states that boundary-layer winds "are backed and decreased from the geostrophic values" S38, with the turning about 30 degrees over land and generally less over water S39. Strategy depends on separating the parts of this structure that repeat, which can be planned for, from the parts that are single events, which can only be exploited when seen.
12.2.1Oscillating shifts, persistent shifts, and mixed patterns. Oscillating shifts swing around a mean direction and come from vertical instability; persistent shifts move one way and come from fronts, or shoreline effects S52. Bethwaite's measurements, reported by SailZing, give directional with amplitudes from 5 degrees in light-medium air to 40 degrees in heavier air S50. The decision rule differs by type: in an oscillation the boat tacks on headers to stay on the lifted tack, in a persistent shift it sails toward the side closer to the expected direction S51. A mixed pattern is an oscillation superimposed on a drifting mean, and the crew tracks both: headers are taken only when they exceed the drift, and the drift sets the side.
12.2.2Pressure bands, cloud effects, shoreline influences, and anomalies. Pressure is uneven at every scale: Bethwaite describes a surface roughly , with gusts carrying 30 to 40 percent extra speed S50. A that is not raining reduces the wind as it approaches, gives null inflow overhead and stronger wind once past S44, with Carlson's lefty on the left side of the cloud and righty on the right side looking upwind S41. Friction "causes the wind to shift left over land in the Northern Hemisphere, which often leads to " S40, so a windward shore produces a band of altered direction and pressure that the model treats as fixed, not as a shift.
12.2.3Sustained pressure feature versus temporary gust line. A gust line is an instability event: Bethwaite's data show wind speed rising and falling by about 7 percent every 6 to 12 seconds S50, and a single gust is exploited for speed, not steered for. A sustained feature has a mechanism that keeps it in place: a sea-breeze front, which fills first offshore and moves toward land S51; a convergence band near a windward shore S40; or the stronger wind that Bethwaite advises scanning for at , at least 200 to 500 metres away S50. The test is persistence over more than one gust cycle and a boundary that moves in a predictable direction. Only sustained features earn a change of side.
12.2.4Location and movement of transition zones. A transition zone is the boundary between two wind regimes: the edge of the sea breeze against the gradient wind, the edge of a cloud's inflow, or the seam where a shoreline bend meets the open-water direction. A sea breeze in the Northern Hemisphere rotates clockwise through the day S42 and Dellenbaugh estimates the movement at roughly 5 degrees per hour S52, so the sea-breeze side of the zone changes direction on a known schedule, while the rate at which the boundary itself moves has to be timed on the water. The crew locates the zone by the line between two water textures and by boats on opposite sides holding different headings, then plots which leg will cross it and on which tack the crossing gains.
12.2.5Tracking changes through fleet heading, heel, and relative speed. Other boats are anemometers distributed across the course. A boat ahead on the same tack that heads up or bears away reports a shift arriving before it reaches the crew; a boat that heels or sits up reports pressure; a boat that gains bearing without a heading change reports either pressure or a favourable current gradient. Dellenbaugh's geometry converts these signals into distance: a 5-degree shift changes the relative position of two boats by 12 percent of their lateral separation S52. The crew reads the fleet in a fixed sequence, nearest boats for shift, distant boats for pressure, so that no signal is missed and none is double-counted.
12.2.6Gold-medal standard: identify the mechanism, then anticipate movement rather than react. An observer with the tracking replay sees the boat tack or gybe before the fleet around it changes heading, on features that were named on the pre-start record: the sea-breeze edge at the time the model predicted, the right-hand side of the cloud, the convergence band under the shore. Tacks made after a header has already cost distance are counted and trend downward through the season. Asked after the race to name the mechanism behind each major shift, the crew gives the same account the wind record supports.
12.3Current and tidal strategy
Current moves the whole racecourse, the boat, the fleet and the water the marks are anchored in, and unlike wind it is largely predictable from a table. Its effect enters every geometric calculation: the bias of the line, the layline to the windward mark, the approach to the gate, and the apparent heading changes that can be mistaken for shifts. The German federation's coach syllabus lists tidal atlases and tide tables, current recognition and measurement as required knowledge S70; this section states how those inputs become decisions.
12.3.1Establishing current direction, strength, and course variation. The tide table gives the prediction; the water gives the fact. Saxton reads current from "lobster pots, moored boats, or even the choppiness of the water" S68, and Jobson from "current ribbons, trap buoys, birds and fish, and the fishermen who congregate at tidal rips" S69. The crew checks the wake on the committee boat's anchor line and on any mark before the sequence, then compares the lean at the pin with the lean at the boat: a difference in strength or direction across the line is the first evidence of a gradient. Direction and strength are recorded with the time, so that the observed turn can be compared with the tabled turn.
12.3.2Effects on line bias, starting distance, and acceleration timing. Current shifts the effective bias: a set toward one end carries the fleet toward it and makes that end easier to reach, while a set along the line makes the up-current end effectively longer to sail to. Adverse current lengthens the time to cover the final distance to the line and demands an earlier commitment; favourable current shortens it and demands a later one. The 2015 49er Europeans tracking showed a of 10 m, with the starboard-tack advantage under one second in time to the line after the gun, measured on speed over ground S16; that gap is recomputed for the current the crew has measured, not the current the table predicted.
12.3.3Effects on windward laylines and leeward-gate approach. When current sets toward the windward mark a boat tacking on the wind-only layline overstands, and in adverse current Saxton's advice is to "take your medicine early and tack away again while you're in the weaker current" S68. Perry shows that a boat cannot gain by pinching to put the current on its lee bow S67, so , not by pointing. At the gate the down-current mark turns faster S58, and a cross-set moves the layline to each gate mark in opposite directions: the crew aims for the mark the current is carrying it toward, unless the hoist or the next beat's side argues otherwise.
12.3.4Separating current-induced heading changes from wind-shift evidence. Current changes the boat's course over the ground without changing its heading through the water; a wind shift changes the heading. The compass therefore answers the question directly: a bearing change to a mark with a steady compass heading is current, while a heading change with unchanged sail trim and apparent wind angle is a shift. Boat-on-boat evidence is ambiguous, because a boat in stronger favourable current gains bearing exactly as a boat in a lift does. The crew therefore resolves any apparent shift against the compass and the mark bearing before acting on it, and treats gains that appear only on the ground track as current.
12.3.5Using geography to anticipate current gradients. Current strength follows the shape of the channel and the depth: it runs stronger in and weaker over shallows, accelerates past headlands and bends around them, and the turn arrives earlier close to shore than in the main stream. Houghton and Campbell's venue treatment combines bends, bands and shifts near shores with topography S43, and the same map serves for the current. Dellenbaugh notes that "If , the water is smoother than when the current runs against the wind" S51, so the water texture itself maps the gradient: a band of steeper chop marks where the current opposes the wind most strongly.
12.3.6Gold-medal standard: include current in major tactical calculations. Observed on the tracking, the boat's layline calls, gate choices and final approach to the line each carry a current correction that matches the measured set, and the crew can state the correction it applied. Laylines are hit rather than overstood or under-laid in a cross-set, the down-current gate mark is chosen when the set favours it and the least crowded mark otherwise S58, and no tack is called on a heading change that the compass shows to be a change in course over the ground rather than a change in wind.
12.4Wave-field strategy
Wave physics is fixed by wind speed, duration and , by depth and by current, so the wave field on a racecourse can be mapped before the start and predicted through the day. Waves grow until the sea is fully developed for the wind that made it S71, and swell is wave energy that has travelled out of its generating area S71. The crew treats waves as a third field alongside wind and current: usually the least valuable of the three, but decisive when the other two are even.
12.4.1Mapping wave direction relative to each leg. Wind waves align with the wind that made them, but leftover waves can run at an angle to a new breeze, so that one tack is into the waves and the other across them S77. The crew records the angle between the primary wave train and the wind before the start and derives the angle per tack: upwind, the tack heading closer into the waves pounds and loses; downwind, the gybe running closer to along the waves surfs longer. Where swell and wind waves cross, the crew notes the two directions separately, because each leg meets a different combination of the two.
12.4.2Areas of better wave shape, spacing, and surfing potential. In , where the depth exceeds half the wavelength, phase speed is about 1.56 times the period in metres per second S73; from a depth of half the wavelength waves , slow, shorten and steepen, and their crests refract toward the shoreline S72. raises wave height and sea disturbance above normal; wind with current lowers them S76. These three effects locate the surfable water: longer, faster waves offshore in deep water, steeper and shorter waves near a shoal, and the steepest chop where current opposes wind. A wave breaks when its steepness exceeds about 1:7 S74, which sets the limit of usable steepness.
12.4.3Tactical cost of routing toward better waves. Sailing to better water is a lateral separation from the fleet, and lateral separation is priced by the wind: a 5-degree shift changes the relative position of two boats by 12 percent of their lateral separation S52. The gain from waves must therefore exceed the expected loss from shifts over the distance travelled, and the percentage principle of many little gains rather than one big gain S51 applies. Downwind, where Dellenbaugh places the biggest gains and losses in differences of wind speed across the course S51, the wave route is taken only when it coincides with pressure or with the favoured gybe.
12.4.4Wave conditions and mode selection. Each tack and each gybe carries its own mode when the wave angle differs S77. Upwind into steep waves the boat needs more depth and twist for acceleration through each face, and a lower heading; across waves it can point. Downwind in surfable waves the boat bears away on the face and heads up on the back, so that the mean course stays on the VMG angle while the instantaneous heading oscillates around it. In extreme conditions Bithell notes a pitchpole risk when the boat ploughs into the back of a wave at full speed, and the gennaker is oversheeted to slow it S29. Mode changes at each tack are part of the plan.
12.4.5Anticipating sea-state changes through the race day. Waves lag the wind: they grow with duration and fetch S71, so a building breeze produces a sea that keeps growing after the wind steadies, and a dying breeze leaves waves running long after the pressure has gone. A sea breeze that rotates clockwise through the day S42 leaves the earlier wave train running at an angle to the new direction S77. A tide turn that brings current against the wind steepens the chop S76. Significant wave height is the mean of the highest third and the largest individual wave is about twice that figure S75, so the crew plans its mode for the largest wave, not the average.
12.4.6Gold-medal standard: integrate wave opportunity without abandoning higher-value wind or current information. The observer sees the boat's mode change at every tack and gybe on a day with an angled sea, and sees no lateral excursion toward better water that is not also a move toward pressure, a favourable shift or a current gain. Debrief notes for each race record the wave angle per tack and the mode chosen, and the wave route is credited only where the tracking shows the gain exceeded the shift cost. Where wind and current are even, the wave map decides the side.
12.5Course geometry and race management
The course is a set of fixed points on a moving medium, and race management adds its own variables: line length, offset marks, gate width, and the option under rule 32 to shorten or abandon after the start S1. Geometry determines where the fleet will crowd, where laylines converge and where a lead can be lost without a wind change. This section links geometry to the model so the crew knows, before the start, where the race is expected to be decided and where it can be reopened.
12.5.1Start-line bias relative to true wind, current, and preferred side. The favoured end is found by comparing the bearing of the line with the head-to-wind bearing, and for every 5 degrees of upwind bias to one end that end is closer to the windward mark by about one eighth of the line length in distance sailed S65. Bias is measured at the same time as the wind direction, because an oscillation moves the favoured end through the sequence. The current correction from 12.3.2 is added, and the preferred side then ranks above both: Dellenbaugh's rule is that line bias is overridden when one side of the course is strongly favoured S51. A transit through both ends to a fixed object ashore locates the line S51, Nick Craig using two or three S66.
12.5.2Course symmetry, unequal leg length, and offset-mark implications. A windward mark set off the wind axis makes one tack longer than the other, so one side of the beat has more room to play shifts and the other reaches its layline sooner. The crew measures the mark bearing against the mean wind during the sequence and records which tack is long. An offset mark forces a short reach before the hoist, so the last approach to the windward mark is on the tack that gives a clean rounding and a set without traffic, and the rounding is executed with the boat close-hauled before the bow reaches the mark S51. Leg length sets how many oscillation cycles the leg contains, from the periods in 12.1.4.
12.5.3Gate separation, gate bias, and likely traffic distribution. At a gate the mark farther upwind is closer, the least crowded mark is preferred and the down-current mark turns faster S58; a pinwheel at the right-hand gate mark can pin a boat indefinitely S59. A wide gate turns the last part of the run into a choice of sides for the next beat; a narrow gate reduces the choice to traffic. The crew predicts the traffic from the fleet's position on the run: the mark on the side the leaders favour will be crowded, and the clear rounding at the other mark is often worth more than the length gained at the favoured one.
12.5.4Locations where fleet compression is likely. Compression occurs wherever laylines converge: the windward mark, the gate marks, and the three-hull-length zone S1 in which mark-room is decided. It also occurs at any boundary of the course and at the edge of a that reaches the leaders first and holds them while the fleet closes. Dellenbaugh's rule to "Stay away from the corners and laylines" S51 reflects that a boat on a layline has no options left and every boat behind it can still gain. The crew maps the compression points before the start and plans to arrive at each with the inside position, which is worth fighting for S51.
12.5.5 Geographic position, clear air, or : choosing priority. The three cannot all be held. Geographic position means being on the favoured side; clear air means a lane free of the wind shadows that Baird places within 6 to 10 boat lengths upwind, the greater distance in lighter air S56; flexibility means staying able to tack on the next shift. Walker's categories decide the order S62: when one side is obviously advantaged, position outranks the others; in an oscillating wind, flexibility does, because a boat pinned in a lane cannot tack on the header; in the third category, clear air and consolidation outrank both. The priority is declared before the start and changes only when the category changes.
12.5.6Gold-medal standard: know where the race is likely won, lost, compressed, or reopened. Before the warning signal the crew can name the leg and location where the model expects the largest gain, the point at which a lead would be most exposed, each compression point, and the feature, a gate choice or a sea-breeze arrival, that could reopen the race. On the tracking, the boat's route shows it arriving at each compression point on the tack and side it predicted, and its , "the extra distance sailed upwind from course middle line on either side" S17, is consistent with the declared category rather than with a corner.
12.6Information hierarchy
Information reaches the crew from forecasts, from the coach, from its own instruments and eyes, and from the fleet, and each source has a resolution, a delay and a bias. The hierarchy decides which source wins when they disagree, and the answer depends on the timescale of the decision: a forecast shapes the model, an observation shapes the next tack. Without a declared hierarchy the crew defaults to the loudest or most recent input, which is rarely the best one.
12.6.1Direct observation versus forecast information. Forecast models resolve what : the ECMWF global model runs at 9 km S45 and GFS at about 13 km S46, while AROME runs at 1.3 km S47 and HARMONIE at 2.5 km S48, and kilometre-scale models represent sea breezes better S49. A racecourse is a small fraction of a global grid cell, so the forecast gives the regime, the timing of a change and the gradient direction, and observation gives the local bend, the exact time and the pressure distribution. The rule is that and the observation updates it; an observation that contradicts the forecast wins on the water, and the disagreement is logged for the review.
12.6.2 against fleet behaviour. The boat's own compass and apparent wind are precise but local; the fleet is distributed across the course but each boat is also responding to its own tactical situation. A single boat heading up may be in a lift or may be pinching to clear a lane, so the fleet is read in groups: several boats on one side heading the same way is a shift, . The crew's own observation is weighted highest for the next tack and the fleet's behaviour highest for the side, because the fleet samples regions the boat cannot reach. Where the two disagree, the boat trusts its compass for the shift and the fleet for the pressure.
12.6.3Using coach input without dependency. The coach sees the course from a stable platform, with the forecast, the tide and the whole fleet in view, and gives that picture up to the warning signal; the Paris 2024 Notice of Race prohibited voice, radio or data transmissions from the first warning signal to the end of the last race of the day S4, so after that point the crew races on its own model. The coach input is therefore fed into the model as a ranked observation , not as an instruction, and the crew rehearses producing the model without it on training days so that the picture remains its own.
12.6.4 Identifying and confirmation bias. Confirmation bias appears as a model that only accumulates supporting evidence: headers on the favoured side are called and headers on the other side are called persistent. The defence is the invalidation signal from 12.1.4, named before the race, so that the disproving observation has a slot waiting for it. Misleading information includes a heading change caused by current rather than wind (12.3.4), a boat gaining through a speed edge rather than a shift, and a gust line mistaken for a sustained feature (12.2.3). Each has a test, and the crew applies the test before the information is allowed to change the plan.
12.6.5Communicating observations without overloading the crew. World Sailing describes the 49er division of labour as the helm making the tactical decisions and steering while the crew undertakes more of the physical work and controls the sails S9, and Sail Canada's progression trains the crew to initiate tactical decisions upwind and the helm downwind S81. Observations therefore flow toward whoever is initiating on that leg, in the vocabulary and the crew has already defined in its own protocol. An observation is spoken only if it changes a decision or updates a named uncertainty; everything else is held until the leg ends. Silence is the default, and a call is a claim with a confidence attached.
12.6.6Gold-medal standard: prioritise information that changes the next decision. On a recording of the crew's communication, every call between the warning signal and the finish can be matched to a decision it informed or an uncertainty it resolved, and calls that did neither are absent. When forecast, coach, fleet and own-boat inputs disagree, the crew can state afterwards which source it weighted and why, and the choice follows the declared hierarchy rather than recency. Observations that contradicted the model were acted on within the cycle in which they were seen, and are logged for the review.
12.7Strategic adaptation during racing
The model built before the start is expected to be wrong in detail and sometimes wrong in kind, and the race is the test. Adaptation is the disciplined replacement of one hypothesis by another when the evidence crosses a threshold set in advance, not a continuous drift with every gust. The 49erFX data study found that VMG on the first beat had a "almost ten times that of any other skill" S37, and the first beat is the largest source of evidence.
12.7.1Updating the racecourse model after the first beat. The first beat gives a full sample: the actual range of headings, the actual period, which side gained and by what mechanism, and where the pressure sat. At the windward mark the crew states the revised model in the same form as the pre-race one, favoured side, mechanism, , uncertainty, and checks it against the course geometry for the run and the next beat. A side that gained through a persistent rotation stays favoured only if the rotation continues; a side that gained through pressure stays favoured only if the pressure has a mechanism that holds it. A side that gained by chance is not carried forward.
12.7.2 for reversing the initial plan. Reversal costs the lateral separation that the distance already sailed toward the original side has built, which a shift then prices at 12 percent per 5 degrees S52, so the threshold rises with the distance committed. The rule is that the plan is reversed when the named has appeared and persisted for more than one oscillation cycle, or when a sustained feature (12.2.3) is confirmed on the other side, and not on the strength of or a single boat's gain. Below that threshold the boat stays in its category and continues the plan; above it, the reversal is executed at once, on the next convenient shift.
12.7.3Managing uncertainty when observations conflict. When the compass says lift and the fleet says the other side is gaining, or the forecast says rotation and the water says oscillation, the model drops to Walker's third category, everything else, and the boat sails conservatively toward the middle and consolidates S62. This is a deliberate reduction in exposure while the crew gathers the observation that will resolve the conflict, and it follows the percentage principle of many little gains rather than one big gain S51. The crew names the observation it is waiting for and the time by which it expects it, so that the conservative mode has an exit condition.
12.7.4Avoiding attachment to an invalidated strategy. Attachment shows up as reasons invented after the fact for staying on the original side. The defence is structural: the invalidation signal was written down before the race, and when it appears the plan is already dead by the crew's own rule, so the decision to leave is not being taken under pressure but was taken ashore. Dellenbaugh's warning that laylines "are definitely a dead end when the wind is shifting" S51 applies to strategies as well: a plan that has reached its layline has no further options, and the remaining choice is only how much to lose before the mark. The crew treats a reversal as execution of the model, not as its failure.
12.7.5Re-establishing priorities after a poor start, penalty, or loss. After a poor start the boat is in bad air, where Baird's data place the exit within 6 to 10 boat lengths upwind, the greater distance in lighter air S56, so the first priority is a clear lane on the tack closer to the favoured side, and the model is unchanged. After a penalty turn the priority is the same, plus the loss of the surrounding boats as reference. After a large loss the temptation is a flyer; Dellenbaugh's principle treats a first-beat flyer as a lottery ticket S51, and the series score, not the race, decides whether that ticket is worth buying. The plan is reissued in one sentence, then executed.
12.7.6Gold-medal standard: evolve strategy with evidence while execution stays decisive and calm. On the tracking, changes of side after the start each coincide with a logged invalidation signal or a confirmed sustained feature, and no change of side follows a single header or a single boat crossing. The communication recording shows the revised model stated at each windward mark in the pre-race form, and the same tone of voice and vocabulary after a penalty as before the start. Manoeuvre count and quality after a setback match those of a clean race, which is the observable evidence that the reset was completed.
12.8Strategic review and learning system
The review closes the loop between prediction and outcome. A race is a test of the model, and a test whose result is not recorded teaches nothing. The system has a fixed form: the prediction is written before the start, the outcome is recorded from tracking and the wind log after the finish, the difference is classified by the type of error, and the classification produces the next race's observation plan. The output is a venue and mechanism library the crew can consult before every regatta.
12.8.1Recording predictions before racing. The prediction is the model of 12.1 written in a fixed template: favoured side, mechanism, , main uncertainty, , expected wind range and period, expected current at the start and the turn time, wave angle per tack, line bias and start position. The forecast source and its resolution are recorded alongside, so that a forecast error can be attributed to the model that made it. The record is made before the start, not reconstructed after, because a reconstructed prediction absorbs the outcome. The crew's own logged values are the reference; where the published record is silent, no figure is entered and the field is marked crew value.
12.8.2Comparing expected versus actual racecourse behaviour. After racing the crew reconstructs the actual wind field from its compass log and the tracking, which publishes per-boat gap to leader, average speed over ground, distance travelled, number of manoeuvres, and S78. Where a logger is permitted under S3, the Vakaros Atlas 2 reports shift tracking and VMG S79 and the Sailmon MAX records and line bias S80. The comparison is made item by item against the template: whether the side paid, whether the mechanism operated, whether the shift arrived at the predicted time and whether the current turned when the table said it would.
12.8.3Errors of observation, interpretation, timing, or execution. Every difference between prediction and outcome is assigned to one of four classes. An observation error means the signal was present and not seen: the sea-breeze line was visible and unrecorded. An interpretation error means the signal was seen and assigned the wrong mechanism: a persistent shift read as an oscillation. A timing error means the mechanism was right but the clock was wrong: the breeze arrived one race later. An execution error means the decision was right and the boat did not deliver it: a late tack, a slow set. The classes need different remedies, and mixing them produces the wrong training plan.
12.8.4 Venue-specific knowledge . A venue file records mechanisms, not outcomes: the shoreline bend and the conditions in which it operates, the sea-breeze arrival window against the gradient direction and cloud cover, the current gradient against the tide phase. Houghton and Campbell's treatment of bends, bands and shifts near shores, land and sea breezes, water temperature, topography and clouds S43 gives the categories. A rule that says the left paid on a given date is not a rule; a rule that says the left pays when the gradient is offshore and the sea breeze fills is one, and it is confirmed only when it has held on more than one day with the same inputs.
12.8.5Turning debrief conclusions into the next observation plan. Each error class points at a task. An observation error adds a scan to the pre-start routine, at the range Bethwaite prescribes of S50. An interpretation error adds a test, such as the compass-against-bearing check of 12.3.4, to the point in the sequence where the misread occurred. A timing error changes the uncertainty named in 12.1.2 for the next race at the same venue. An execution error is passed to the boat-handling programme. The next race's template then carries the new item, and the review checks whether it was done.
12.8.6Gold-medal standard: each race improves the team’s predictive model. Over a regatta, the record shows the pre-race predictions of favoured side and mechanism converging with the observed outcome from race to race, with the confidence figure the crew assigned tracking its actual hit rate. Every race has a prediction written before the start and a classified comparison after it, no comparison is missing, and no prediction has been edited after the finish. The venue file for each regatta contains at least one mechanism statement that was tested on more than one day, and the observation plan for the next event contains the items the last debrief generated.