MeetRhter

I watched F1 since about 7 years old, and rarely missed a race since 2001. While my affliction for Ferrari's red cars still lingers, with every passing year I consider myself more a fan of the sport in general.
After losing my first F1 Fantasy season in 2019, I've set out to do better than Excel next time around. As COVID impacted the entire world, I found myself stuck at home with a lot more time on my hands, leading to a perfect storm of enabling to do something a bit more significant in F1 Fantasy. I've invested time into building a model to estimate F1 car performance first, and to estimate Fantasy points second.

Starting off with the exceedingly obvious, any Fantasy game is won by assembling a team that scores the most points. The main question then becomes - how do you go about shaping your team that is best positioned to do just that? Perhaps the easiest way would be to look at the Fantasy points scored within the last few races and take an average. While certainly easy, F1 seasons are relatively short in terms of number of races as data points and any averaging--no matter how careful--is sure to include outliers that might be unrepresentative of the underlying performance. Said differently, if you were to rerun the race(s) with slightly different conditions, you could reasonably expect to see different final finishing orders. We can do better than after-the-fact averages. Below, I will describe the current state of the model and the parameters used to estimate expected Fantasy points.

Each driver has the following parameters: qualification pace (Qpace), race pace (Rpace), dnf probability (DNF%), fast lap probability (FL%), and grid penalty amount. Beyond those primary ones which are based on some sort of an objective calculation, there are also subjective parameters like team-order probabilities, and DNQ estimates as a consequence of known grid penalties.

Qpace and Rpace are by far the most consequential. Qpace is a weighted combination of previous qualification results as well as FP1/FP2/FP3 results. Rpace is based on Qpace and typical number of overtakes. Any confirmed grid penalties will detune Rpace commensurate with perceived overtake difficulty. Figure 1 is showing a typical simulation run. Given means and standard deviations for each driver, the effective probability density functions are sampled in a Monte-Carlo-esque way a large number of times (typically N = 10,000). Some additional parameters affect those runs. DNF and team-order probabilities get sampled over those N runs. Fast lap is determined as a combination of derived Rpace as well as the historical (in-season) fast laps.

Each one of the N simulations is then scored consistent with F1 Fantasy rules. There is a number of ways to visualize this data. On Discord, we will be posting the violin plots (sideways histograms), as well as tables with the overall averages. The dashboards are shown in Figures 2 and 3, while Table 1 shows the summary of the simulation run. At the moment, these averages are uploaded and made available within the team calculator. We do have plans to further enhance the experience with additional user-based adjustments; stay tuned.

While I have been a contributing member of the F1 Fantasy Community for almost three years, most of my interest lies in the figuring out of the underlying car performance. Calculating best teams, determining when to use MD, and keeping track of the particulars of the Fantasy game is necessary but can be annoying. This is where I believe joining of Fantasy Tools will be the most beneficial in the long term. We have brainstormed a great many continuations of how to improve both the accuracy of our points estimations as well as the usefulness of our optimal team calculations. As a long-term goal, I intend to improve the model in a number of ways, the most significant of which are: a) evaluating track-similarity for additional result weighting, and b) estimating car characteristics to predict each constructor's performance at a given track layout. Partnering with Fantasy Tools will allow everyone to focus exactly on where their passions lie. We hope that you find what we offer useful.