Keeneland: How Winners Won Races (Race Evolution)

In the spirit, of Keeneland’s opening week I wanted to take a look at running styles and how they fare on various track configurations at Keeneland.

I hope that this can help you think through race design in showing how races have been run, and result in you padding the bankroll a bit ahead of the derby.

Early reflections on the last article

Just a couple of points on ROI and Impact Values, as was brought up in the Keeneland Select Players’ Podcast. If you missed that article you can find it (https://inthemoneypodcast.com/2019/04/02/keeneland-looking-at-the-spring-meet-through-data/).

1. A negative impact value should really only suggest that an angle plays out as successful less than the public expects it to, based on the odds that a horse broke from the gate at. If a trainer wins at 30%, but their odds suggest they should win at 40% (i.e. horses consistently leave the gate at low prices), there would be an impact value of < 1.
2. A negative ROI, doesn’t speak the the percentage of time that an angle plays out at, and shouldn’t result in you leaving horses off of horizontal tickets, though it might lead you to bypass a win bet (I bet primarily to win, and my sentiments in the previous article were based on my inclination for win bets). Ultimately, it should be taken as answering the question: if I consistently play this angle, how much money should I expect to make (or lose)? The odds a horse leaves the gate at play a lot into this and negative ROIs could be indicative of horses going off at very short prices, and not winning enough to generate a profit. To underline this point, let’s imagine a scenario where three horses leave the starting gate at even money for a trainer, and that we are measuring that trainer’s ROI. In order for this trainer to have a positive return, they would have to win at least twice (returning a profit of 2 dollars), suggesting they would need a 66% win rate on 3 dollars invested suggesting a $0.66 ROI (2/3). In another scenario, consider a trainer running three horses that leave the starting gate at 4 to 1. If they win 1 and you bet them 3 times you’ll return 4 dollars, and you’ll have invested three dollars, so your ROI is $1.33 (4/3). In each case the trainer won at 33%, but the prices that their horses went off greatly influenced the ROI of the angle.
3. Lastly, to highlight the point in the podcast: let the price dictate your play. The stats are just aggregations of how events have played out in the past. The future is not necessarily going to follow the same path as the past, but nonetheless can provide at least some indication of how we might expect it to play out with no new information.

Analyzing how winners won races

When it comes to how races evolved, we took a look at races in the following fashion.

• First we broke the races into the different types, much like we did in the last article. The four different types are: dirt around one turn, dirt around two turns, turf around one turn, and finally turf around two turns.
• Then, we look at runners at specific points in the race, based on the distance.
• Then we labeled where the runner was relative to the pacesetter (e.g. were they on the lead, or how far back, measured in lengths, were they from the leader?)
• Then we looked at if they won, placed second, or third, and compared that to all of the other runners that were in the same situation that they were in at each point in question.
• Impact value is calculated as the amount of times a runner actually won, divided by what percentage of time their odds suggested they would win.
• In no way should this data suggest that horses can’t win from being in any position on the track.
• This part is important: this does not look at how horses were expected to run, but only how they actually ran over the various configurations we’ve selected below

The following are the results.

Dirt, one turn

Early race

1/4 mile pole

Since the dirt track was installed in 2014, there have bee around 650 one turn dirt races run. Of those, here are the finishing percentages of each of the running buckets that I have laid out, in the early going. Rationale for the running buckets, briefly, is that I have designated the leader, someone “breathing down the leader’s neck” and then roughly each fifth of a second slower than the leader as a bucket (1-2 lengths, 2-3 lengths, etc.) through more than 5 lengths back. This should hopefully help people translate past sectionals into where they think the horse in question would end up in the below chart, and get a feel for how often that position went on to win in the past.

As most people probably expected, speed plays well in the shorter distance dirt races.

What I found most interesting here is the percentage of horses that led at the quarter pole who ultimately ended up winning the race (over 28%) and the impact value of 1.6 suggests that there is a decent amount of value to find longer priced horses who might find themselves on the lead in the early going, and end up winning the race. They win much more frequently than their odds suggest they should.

Late race

• 1/2 mile for races 6 or more furlongs
• 3/8 mile for races < 6 furlongs

Note here that some of the shorter races will fall off here due to the way that my database is set up, so the number of winners won’t add up to the above. I’ve also added a bucket of more than 20 lengths because the race is more evolved at this point, and more than 20 lengths may suggest the horse is being pulled up. This is consistent for all late race data tables.

• Not too much new here. Races still have some distance left to be run, but they are often decided by the front runners (in this case almost 32% of the time).
• Best thing to take away here is that you can root with some confidence, 32% of the time, that the pacesetter you bet will ultimately win after this call.
• My takeaway, is that given the pacesetter still doesn’t win 68% of the time, maybe save some of your cheering until the finish line.

If you bet to win and place, 50% of the time you can start walking to the counter if the pacesetter you bet is on the lead at this stage of the race

Dirt, two turns

Early race

• 1/4 mile for races less than 10 furlongs
• 1/2 mile for races 10 or more furlongs

The table below shows the evolution of the race in the early going for dirt races run around two turns. For this race configuration, there were about 380 races run, again since the dirt track was installed in 2014

Again, as expected, speed and position is important. However, it’s not nearly as important as in the one turn races. This could be a product of different running tactics after breaking from the gate (think of a runner using speed from an outside post to gain early position, but using too much energy to finish strongly).

• It’s still pretty cool to see that if you predicted the pacesetter, you had over a 50% chance of being in the money.
• I think it’s also very interesting to see the win percentage consistently drop off by how far from the pacesetter the runner was.
• Note that the impact values of buckets 4 through 7 suggest that the amount of money horses that are off the lead take is probably more than they merit taking.

Late race

• 1/2 mile for races less than 10 furlongs
• 3/4 for races 10 furlongs
• 1 mile for races at 12 or 13 furlongs
• 1 1/4 mile point for races at 14 furlongs

Honestly, I am not going to point out much here as there isn’t much to take away. The table above looks very much like the chart for the early race table for dirt races run around two turns

Turf, one turn

Early race

1/4 mile pole

Now we shift to turf races, and repeat the same patterns as above. For one turn turf races, we’ve aggregated about 70 races. The results are below:

• What pops out to me is the stark contrast between dirt one turn races, and turf one turn races. The pace setter here wins about 10 percentage points lower than on dirt (albeit a much smaller sample size).
• That being said, pacesetters still boast an impact value of around 1.5 suggesting there is some value in trying to find and bet (to win) the pacesetter.
• Also interesting to me is that impact values bounce above 1 for runners that find themselves in buckets 5 and 6.

Late race

• 1/2 mile for races 6 or more furlongs
• 3/8 mile for races < 6 furlongs

Not much change here either (relative to the early race for one turn turfs. I’ll let you draw your own conclusions, but for consistency’s sake, I included this.

Turf, two turns

Early race

• 1/4 mile for races less than 10 furlongs
• 1/2 mile for races more than 10 furlongs

Finally we take a look at two turn races on the turf. There were over 400 of such races. Check out the chart below:

• The leader’s win percentage consistently drops off again. It’s about half of what dirt sprints were for reference (28% compared to just over 14.5% here).
• Of note here though is that there does appear to be a “sweet spot.” Take a look at bucket 3 where the win percentage is the highest of all of the buckets, and the impact value is pretty high (> 1.4).

Late race

Finally, we took a look at points late in the race for two turn turf races.

• 1/2 mile for races less than 10 furlongs
• 3/4 for races 10 furlongs
• 1 mile for races at 12 or 13 furlongs
• 1 1/4 mile point for races at 14 furlongs

• That sweet spot we discussed above in bucket three remains the same.
• I would also maintain that closing runners are probably over bet.
• In that same line of thought, I still think that horses expected to be on the lead are underestimated in terms of their ability to win on turf (based on the parimutuel odds).
• I wouldn’t conclude that it’s too hard to close at Keeneland. Asking a horse to close a lot of ground is difficult anywhere, and I would imagine the above chart wouldn’t change too much. That being said, horses that were further out of the running of the race, undoubtedly won a lower percentage of time than those that were “in the race”.

Conclusion

I hope that this was helpful in helping you think how races are typically run, and won at Keeneland. As always, reach out or comment if you have any questions (Twitter handle @john_camardo).