Alright, so lemme tell you about this thing I messed around with the other day – Fresno Fair racing. Sounds kinda simple, right? Well, buckle up, cause it was a bit of a ride.
It all started when I was scrolling through some old forums and saw a thread about simulated horse racing using data from the Fresno Fair. Figured, “Hey, why not? Sounds like a laugh.” So, first thing I did was dig up some historical race data. Man, that took longer than I thought. Found some spreadsheets buried deep in the Fresno Fair website. They were messy, like real messy, you know? Different formats, missing data, the whole shebang.
Next up, I had to wrangle that data into something usable. I’m talking Python and Pandas, baby! I loaded the spreadsheets, cleaned up the inconsistencies, and created a nice, tidy dataframe. Removed all the rows that didn’t have complete information, then converted everything to a more usable format, like changing the times to seconds and cleaning up the jockeys and horse names.
Okay, so now I had clean data. Time to build a basic simulation. I wanted to keep it simple to start. I decided to use the historical data to predict the outcome of a race by assigning a probability to each horse based on its past performance. I tried to use the speed, the jockey, and the historical performance.
Then came the fun part: running the simulations. I wrote a little script that would randomly pick a winner based on those probabilities. I ran it a bunch of times, like hundreds of times, to see what the overall results looked like. It was kinda cool to see which horses were consistently “winning” in the simulations.
But, of course, it wasn’t perfect. The initial results were… well, kinda boring. The same few horses kept winning every time. It didn’t feel very “realistic,” you know? So, I started tweaking things. I added some randomness to the probabilities to simulate unpredictable factors like a horse getting spooked or a jockey making a bad call. I even tried adding in some “track condition” variables, just guessing based on the weather on the day of the race.
After a bunch of trial and error, the simulations started to feel a little more believable. There was more variation in the winners, and it felt less predictable. I mean, it’s still just a simulation, but it was getting closer to capturing the chaos of a real horse race.
Here’s a few bullet points what i did:
- Fetched the data.
- Cleaned the data using pandas.
- Implemented a simulation using python by using historical data.
- Tweaked the variables and added randomness.
Overall, it was a pretty fun project. I learned a lot about data cleaning, simulation, and, surprisingly, horse racing! Would I bet real money on my simulation? Probably not. But it was a good excuse to mess around with some data and try something new.
