The Psychology of the Start Game

Soaring is fundamentally a sport that balances risk mitigation with optimization of performance. At every stage of a flight, a pilot is presented with a risk/reward scenario wherein he may gain time on his competitors. This is offset by the risk the pilot exposes himself to; going faster diminishes range, which in turn lessens the pilot’s sampling opportunities. It seems to me that most people in high quality competitions are quite good at optimization. I think most of the top pilots flying alone over a variety of days would have quite similar performances. However, the risk mitigation that competitions afford radically change the equation that normally applies to individual flying. How to mitigate individual risk relative to the competition is perhaps the most critical skill tested. Furthermore, since the group behavior of pilots in a gaggle can be quite inefficient, this makes it all that much more important.

With inefficiencies come competitive opportunities. These inefficiencies develop due to how racers respond to immediate and possible losses. Why losses? Since soaring performance incorporates the relative performance of all contestants through all the contest days, the results strongly favor consistency. Doing better than the rest is good, but doing worse can be catastrophic. This means that pilots generally need to avoid losing rather than trying “to win”. Only at the very end of a competition does a true “race” develop between the contestants at the very top, at which point most of the competitors have dropped out of the running.

The Start Game

When it comes to starting, it is particularly evident how losses drive behavior and the sort of inefficiencies it will expose.  The two possible losses are driven by timing, which is absolutely critical. Start too late, one might get dropped off the back of the day and land short. Start too early, one risks “getting rolled” by the competition, losing quite a lot of points. This creates a tension as there is a present bias to avoiding losses now and there is eventually an even stronger loss aversion to coming up short when others make it home. It is quite painful for most contestants to be in the cockpit knowing with full certainty that other contestants have beaten them, especially in the first stages of the race. As such, there is a conflict between two losses, the more immediate one of getting rolled versus the possible, catastrophic one of landing short.

As a result, most gaggle starting behavior is driven by the point that pilots start feeling pressured by possibly coming up short. Specifically, each individual contestant has a tipping point that integrates how the day is developing thereby driving a prediction of when the day will fall apart. This incorporates the current and possible future weather conditions and the behavior of other contestants as one can generally fly faster and with less risk with company than alone.

Pilots are generally hesitant starting before the optimal part of the day because this will incur a direct loss in performance. However, once the optimal part of the day rolls around, most will be in a position to start. As more time goes on, more and more will be ready to go, which in turn means that starting incurs a definite loss for the individual. However, the possibility of “getting rolled” means that it is best to wait.

As even more time goes on, it becomes more and more evident that it is getting late. Each individual is starting to get anxious; when is the gaggle going to finally go? Soon nearly everyone is worrying about coming up short at the end of the day. At this point, all of the individuals in the pack are facing a crisis for they are feeling the stress of needing to go with the immediate, painful and the quite salient loss of accepting a loss of points now.

As more and more pilots enter this state of crisis, the gaggle is more and more primed to go. At this point, the whole start gaggle becomes unstable and anyone starting (or making the appearance of such) will motivate the group to go. Some savvy (and more aggressive) pilots will hang on just a couple moments longer to take advantage of the flood going to the gate. But the underlying mechanism driving this is the anxiety of starting late becoming overwhelming enough for everyone that the group goes. Prior to this condition being met, the group is unlikely to start.

There are several factors that affect the group’s decision to start.

1)      Estimated peak/end of the day

If the end of the soaring day is likely to be abrupt, this will motivate the gaggle to start closer to the optimal time. If the soaring day is likely to decay slowly, then the gaggle will probably wait a considerable amount of time. The reason is due to the saliency of coming up short in the task. For instance, if everyone knows that a warm front will enter the task area sharply shutting off convection, it is on the top of every pilot’s mind that the day will die and he must start if he hopes to finish. Since everyone will have a similar mindset, this will make the starting decision much more defined. However, if there is the day will decay slowly, then the starting decision becomes much more ambiguous. It makes it difficult to predict with certainty what speed will be achieved. The result is that it will take longer for pilots to start considering the possible loss of coming up short.

So when it comes to predicting the behavior of the gaggle, you must ask, is the starting decision today clear or ambiguous?

2)      Number and position of previous starters

If an individual or groups started before the main gaggle, it will certainly affect the decision to start as it changes the risk/reward structure. First, if a pilot decides to start soon enough, he may be able to run down the group ahead. However, even if this window of opportunity has passed, the gaggle will also be affected by the number of previous starters. The cost of coming up short increases when each individual in the gaggle knows that there will be others that will make it around. In turn, this will make this risk more salient and likely encourage individuals to start sooner.

The counter scenario is when there are no starters and everyone is playing the start game. In this case, it does not matter how fast each pilot goes, since no one has started. In this case, there is little or no motivation to start unless someone sneaks out or there is a false start.

Lastly, it is also important who is in the list of previous starters. If it includes a list of strong individuals and teams, this puts further weight on their assessment of the day. This will motivate the gaggle to start earlier if possible. If the pilots who left early are not a threat, then they are less likely to influence the decision.

3)      The difficulty of restarting

Here it is useful to note the usefulness of “false starts”. The point at which one will feel an overwhelming need to start will differ for each individual. Some will feel the anxiety associated with the timing of the day sooner than others. If some people “fake” their start, it is likely to get some of those folks to go with them. In turn, this will cause some people to start and in turn will affect the optimal equation for each individual.

However, false starts come with their own risks. The major question is, “how difficult is it to return to a starting position following a start?” If it is possible to return to a starting position relatively easily, one can expect some teams to either fake a start or change their mind shortly on course. This can occur in conditions when there are clouds, especially with cloud-suck. However, if the conditions have a weak top of boundary layer, such as in blue conditions, pilots will be hesitant to get out of the start gaggle and lose their starting position. In turn, this means that it will take longer for the gaggle to start as there will be fewer opportunities to attract people out of the gate.

One should ask, what is the cost of restarting? If the cost of restarting is low, expect more false starts and earlier and more distributed set of starts over the contestants. If the cost of restarting is high, expect a tendency toward a later, mass start.

4)      How important is the gaggle for risk mitigation?

Do the conditions influence a risk mitigation or a performance optimization mindset? If most pilots are in a risk mitigation mindset, then they are much more likely to fly with the pack and avoid taking risk. This includes starting early and even doing false starts. This creates the classic case of waiting for someone else to lead out with its associated inefficiency. These conditions are likely to occur at the start if the first leg is quite difficult. Leading out not only incurs an inefficiency, but also incurs the fear of having other pilots simply sitting and watching you mark. This is the same as a bottleneck forming on course which causes no one wanting to lead out due to the risk of doing so.

In contrast is a situation that encourages an optimization mindset. A classic scenario is one where there are many cloud markers and other gliders are less necessary. Pilots that are in a good position to go on their own when they believe it’s time are more likely to start.

If the risk of landing out early is high, it is likely that the gaggle is likely to wait. This is because no one wants to lead out and take the risk upon themselves.

If there are well marked clouds, this will motivate individuals and teams to start. This will change the risk/reward of starting late to encourage the gaggle to go earlier.

In summary, the group’s decision to start is when most people in the group feel an overwhelming desire to start due to increasing threat of coming up short at the end of the day. As more time goes on, the more likely someone will be driven to start or an action will be perceived as a false start. Once enough go, then the group will be destabilized enough to flood toward the gate. The factors that influence when this will occur include if other contestants had started, how defined (or ambiguous) is the peak/end of the day, the cost of restarting and the importance of the gaggle on the first leg.

The cost-benefits of starting early:

Starting earlier than the pack is an aggressive strategy that carries a very high risk and possible high reward. The risk invested is immediate and the reward is unclear. This is because starting early is highly committing and assumes a contrarian view. Unless you predict a particular inefficiency in the gaggle, it is very difficult to be faster than those who fly with the pack. Furthermore, the pilot exposes himself to much greater risk due to the loss of sampling options afforded by the gaggle.

One can consider to start early if one of two sets of conditions are met:

First set:

1)      You do not need the gaggle over the course of the day. There are no expected major bottlenecks/down shifts during the day that will require the help of others. This means you are flying in an optimization mode as opposed to a risk mitigation mode. This is generally when the conditions throughout the day are relatively consistent and homogeneous.

2)      You believe that when you start you will be able to sneak away without causing a multitude of other pilots to start causing you to get rolled in short order. This means that you are either in a very different position at the gate or you start before the peak enough that pilots feel it is too early. Once at peak, starting is nearly guaranteed to get others to start going with you, thereby ensuring a substantial competitive loss.

3)      The peak/end of the day are ambiguous. If they are ambiguous, it will take longer for pilots to start feeling the end of the day pressuring them to start. This means that it will take longer for some pilots to be pressured to start, thereby causing the pack to deviate further from the “peak” of the day. If there is a threat of the day shutting down at the end, everyone will be in the same mindset. This means it is very unlikely that the separation between you and the pack will be substantial enough to make a clean break.

4)      You can generally fly as fast or faster than the pack on your own. This means that you must either have clear cloud markers, or you are flying with a competent teammate.

All four of these conditions must be met for the pilot to perform at least as well as the gaggle. Most of the time, it is quite difficult for an individual or even a single team can accomplish all of this with the score having an expected value that is greater than simply running with the pack. The gains of picking up markers makes it quite difficult to justify the risk, especially since one must accept some inefficiency to make the start work to begin with. However, there also exists a fifth possible condition that makes starting earlier than the pack a much bigger gain:

5)      A bottleneck may form behind you and the risk of taking advantage of it is acceptable. For instance, if you expect that the conditions are currently good to start and will get worse, this provides an opportunity to make a break away that will stick.

In my opinion, the bottleneck scenario is the most attractive competitive opportunity to start early. That being said, it’s not a sure shot either. You must be really sure that there will be a bottleneck because you are taking a large risk in leading out with few or no markers on the first leg. Furthermore, you should feel confident that this bottleneck is one that is not easily recognized by others. Ie: You need to have excellent weather information.

The cost-benefits of starting late

Starting later than the pack can also be quite an effective strategy, with a high risk/reward. The benefits are largely due to the advantages of having markers on course that show where the lift is organized and its climb rate. This provides a greater degree of certainty, which enhances the degree to which the late-starter can optimize his efficiency throughout the day. Knowing that a thermal is four knots ahead makes it easy to fly MC 4.

The risk of starting late is the possibility of getting dropped off the back of the pack and possibly the back of the day. Not only does the pilot lose the benefit of having markers, but he also must deal with the decaying day without much needed help. As such, starting late results in a fine line of timing and positioning with respect to the gaggle and considering the possibility of getting dropped off.

The best case scenario starting late is when a bottleneck develops on course that slows down the gaggle. This creates the opportunity to merge with the gaggle, gaining a considerable amount of time on everyone else.

The decision criteria for starting late include:

1) The day will decay slowly. This lessens the risk of getting dropped off the back and increases the expected value of the gamble as if one gets dropped off, it should be possible to make it back.

2) You expect a bottleneck to develop on course and it is possible to catch up and pass the pack.

3) The day forms defined, exploitable groups. If it is a cumulus day, this is less the case and makes it less attractive to start quite a bit later.

4) The conditions are consistent on course and you do not need their help in the early part of the race. In essence, a bottleneck is not going to develop behind the pack that will burn you whereas the pack sails past.

Both starting late and early have strategic risks and rewards. Starting later generally allows for incremental gains at the risk of a massive loss. Starting early allows for massive gains at a yet higher risk, with a greater risk of a big loss in either getting rolled or worse, landing out early. Deviating much from group behavior requires good reasoning. It is generally a smarter bet to consistently do a little better than the pack rather than looking for heroic gains.