Game theory is the science of strategy, where the outcome for each participant depends on the actions of all. I am a big believer in Game Theory as it relates to cycling tactics. For example: do you chase down a breakaway (e.g. Alaphilippe at the 2020 World Champs) and risk bringing with you someone that can beat you in a sprint? Or do no work, let the sprinter lead the chase, and hope for second place?
Watching the Zwift Racing League premier divisions and then commentating on APAC region, I realised that most races along the Richmond UCI Reverse course were ending in a big bunch sprint.
We decided at the end of our race to put some game theory to the test through a fakey sprint effort, with one rider (myself) putting in a fake sprint attack with 800m to go. Check out the video:
There is always risk with any strategy employed in a cycling race, but using Game Theory you can minimise the risk or at least know what it is and mitigate it on the fly. With the fakey sprint here were the scenarios:
- Fakey sprint draws out our competition and forces them into a chase, bringing up our teammates who can sprint over the top, our team wins
- Fakey sprint does not fool the competition, they stay with the bunch and ride wheels to put in a maximal sprint at the more opportune time, our team loses
- Fakey sprint does not fool the competition, they stay with the bunch but don’t catch, our team wins
So looking at the above there is a 33% risk that we will lose the race with this strategy, but a 66% chance that we will win the race. Therefore, it was an easy decision to make… go for the fakey sprint.
The alternate is obviously that we don’t fakey sprint at all and take the risk with a bunch sprint at the end. But with 5 of our riders in the top 9 this only equates to a 55% chance of winning, still lower than 66%. I have found that the odds always seem to be better when you are the team making the plays, being proactive rather than reactive.
These strategic decisions are also faster and easier to make the better you know your competition and the more familiar you are with the course. So: do your homework!
What About You?
If your team has pulled off something amazing in a race, or you’ve seen something tactically great, shoot me a message (or comment below) and I’ll do some analysis on why and how it worked.
I love going out on a breakway, but Zwift’s dubledraft discourages any attempt, as a compact group will easily be able to reach a higher speed even when doing more strength than the group. Therefore, it is more advisable to stay in the group and try your chances in the sprint. The races would be more exciting if we didn’t have the dubledraft in my opinion.
The good thing about the Zwift Racing League is that it is team racing, so breakaways and fun tactics are easier when you have teammates who can help you pull it off. Obviously smaller bunches are easier to make these tactics stick.
Why do you call this a fakey sprint? It is an attack in the last kilometre, a time tested tactic in the real world. A fakey sprint would be an early “fake” jump with the hope of drawing out your competitors, and getting them to lead you out.
Hey, we faked a sprint as there was a ‘dummy’ finish line banner on the map that we thought would get our competition confused, which it did, and they pulled up our teammates who rolled over the top. Probably longer than a sprint but not quite a long range attack.
> I have found that the odds always seem to be better when you are the team making the plays, being proactive rather than reactive. Indeed. I wonder if this has something to do with how sprints get registered by other competitors “later” than they would in real life. Typically the avatars onscreen (including you!) react after the initial shot has been fired, so to speak. IRL you would pick up on those hints _before_ the wattage increases, rather than after. Or at least that’s been my experience. By the time you see someone else make a move it’s too… Read more »
Good point, by the time you see the orange numbers in zwift it’s probably too late, but it’s harder to anticipate than in IRL…
Interesting article, and well done on winning! I love the strategic approach, but the nerd in me feels that your thinking here is not completely on point. You seem to assume that each of the three possibilities listed has equal chance of happening, which in my view is not a given. You also assume that if you manage to “draw out the competition”, the chance of winning will be 100%; again, not a given as not everyone will chase the fake sprint. I approached this from another angle: What should the chance of success of my fake sprint (p_success) be… Read more »
Love this!!! I have definitely simplified the math here because it’s hard to do lots of calculations during the race when HR is 170+, but totally agree with your points and calculations. Check out this article, I think you’ll find it interesting, an analysis of strategy in cycling as it relates to game-theory. Please share any similar studies you’ve seen, I am fully geeking out on this!
https://halshs.archives-ouvertes.fr/halshs-01326720/document
Thanks! Yeah, I can imagine these things are difficult to calculate in the heat of the moment 🙂
Hard to argue with success. But I’d be more inclined towards rotating attacks. Real ones, for like the last 3-5km. The group is small enough that blob effects are small, so the other teams would have to chase for real. I’d be surprised if it took more than three attacks for one to stick. This allows you to leverage your numbers advantage rather than going with the crapshoot that is a sprint. This strategy is good if all you care about is the win. I don’t know how this strategy works with the points system in ZRL. Winning at the… Read more »
Good points… because we had 5 in the top 9 and ZRL points go 80(ish) deep and weighted toward top 10 we would more than likely take the win even if not over the line. A good time to take a somewhat risky attack strategy.
You assigned equal probabilities to each scenario which probably isn’t accurate.
Agree, this was definitely calculations on the fly at 170+ HR
You don’t know the chance of success for each outcome which renders this meaningless as a calculation. Sorry if that’s sounds abrupt but it’s the truth.
I really love these tactics videos, please keep them coming! Can I ask what tool you use to do the “telestrating” part of them? Thanks.
Thanks Brent! I use EpicPen as the drawing tool. Play the video on second PC and pull it through to Vmix on first Pc using NDI capture