AustinDillon75
Colt
Day 10 of our run. Today there's a 77% correlation to the market. The Chester card for today attached.
Let's see how we go.
Let's see how we go.
AD's daily selections
- Thread starter AustinDillon75
- Start date
AustinDillon75
Colt
I've asked Gemini for an assessment of how we are doing so far and here it is. Forgive me for it being in dollars.
Based on the data provided (379 valid bets), here are the calculated scoring metrics and a performance assessment.
The performance is currently exceptional. A yield of $13.35\%$ over 379 bets is significantly higher than what is typically expected from a long-term sustainable strategy. The "edge" is coming primarily from the mid-price range (BSP between $3.5$ and $8.0$), where you are seeing significantly more winners than the market prices imply.
How likely is it to continue?
The p-value of $0.099$ is the most important number for sustainability. It indicates that there is a $9.9\%$ chance that these results have been achieved through pure luck (variance) rather than a genuine skill-based edge.
The strategy is performing very well. To increase confidence in its continuation, you should monitor the next 200–300 bets. If the yield remains above $5\%$ and the p-value drops below $0.05$, you can be much more certain that the performance is due to a sustainable edge over the market.
Based on the data provided (379 valid bets), here are the calculated scoring metrics and a performance assessment.
1. Accuracy & Calibration Scores
These metrics evaluate how well the implied probabilities (from the BSP) match the actual outcomes.- Brier Score: $0.1941$
- This measures the mean squared difference between the predicted probability and the actual result. For context, a score of $0.0$ is a perfect prediction, and $0.25$ is essentially "guessing" in a 50/50 scenario. A score of $0.19$ in horse racing is quite strong, indicating that the market prices (and your selection process) have a good predictive relationship with the results.
- Log Loss: $0.5714$
- Log loss penalizes confident wrong predictions more heavily. This value is relatively low, suggesting that the outcomes are generally within the "expected" range of the market prices without too many massive upsets where a high-probability horse failed.
- Chi-Square Statistic: $15.32$
- This is a goodness-of-fit test. With 10 bins of data, a value of $15.32$ (p-value $\approx 0.08$) suggests that while there is some deviation between expected and observed wins, the calibration is reasonably tight. However, the deviations we do see are in your favor (observed wins exceeding expected wins in specific price brackets).
2. Performance Summary
| Metric | Value |
| Total Bets | 379 |
| Returns (Yield) | $13.35\%$ |
| Actual Strike Rate | $32.45\%$ |
| Expected Strike Rate | $22.70\%$ |
| t-statistic | $1.29$ |
| p-value | $0.099$ |
3. Assessment & Sustainability
How well is it doing?The performance is currently exceptional. A yield of $13.35\%$ over 379 bets is significantly higher than what is typically expected from a long-term sustainable strategy. The "edge" is coming primarily from the mid-price range (BSP between $3.5$ and $8.0$), where you are seeing significantly more winners than the market prices imply.
How likely is it to continue?
The p-value of $0.099$ is the most important number for sustainability. It indicates that there is a $9.9\%$ chance that these results have been achieved through pure luck (variance) rather than a genuine skill-based edge.
- The Positive: You are beating the expected strike rate by nearly $10\%$. The cumulative PNL curve shows a consistent upward trajectory rather than being driven by one or two "lucky" longshot winners.
- The Caution: In statistical testing, we generally look for a p-value below $0.05$ to declare a strategy "proven." At $0.099$, you are in a "grey area." It is very likely you have an edge, but 379 bets is still a relatively small sample for horse racing.
The strategy is performing very well. To increase confidence in its continuation, you should monitor the next 200–300 bets. If the yield remains above $5\%$ and the p-value drops below $0.05$, you can be much more certain that the performance is due to a sustainable edge over the market.
AustinDillon75
Colt
That was a very attritional day and hard fought day of winning and losing. No losing run of more than 4, but only 2 instances of back to back winners.
Yesterday was 33 races, 12 winners and a BSP profit of 0.53pts. The biggest profit was actually Adonius in the first race of the day, a 4.02, and exasperatingly, our stake on Maillot Blanc was halved due to him finishing in a dead heat for first. Still, we crack on.
TOP RATEDS AFTER DAY 10
393 RUNS
125 WINS
31.8% STRIKE RATE
41.29 PTS TO BSP
10.7% RETURN ON INVESTMENT
I am also recording the top rated horses finishing in the frame too, we've had 72 seconds and 51 thirds, so it's trapping the first two 50.1% of the time and the first three 63.1% of the time. That is pretty good, after 10 days. We just need a couple of bigger priced winners for lift off, we are due and of 37 that went off double digits to BSP we've had 5 second place ones. It just needs a couple of them to become winners now.
If we average all the 10 day periods of the backtesting, this is in the 43rd percentile for profit (49pts the middle), the 60th percentile for strike rate and the 38th percentile for return on investment (55pts the middle). So you can see how a couple of decent winners will make all the difference - they are almost certainly coming.
Note I'm taking the value bets off for now for time reasons and also because it's flagging anything the first 3 of the ratings no matter the number of runners. It needs adjusting so that it can only be a pick if the each way terms would justify it. I don't think it's right to make the 3rd ranked as a value pick if there are only 4 runners.
So today's picks attached and the Chester racecard. An 80.2% market correlation with morning lines.
Yesterday was 33 races, 12 winners and a BSP profit of 0.53pts. The biggest profit was actually Adonius in the first race of the day, a 4.02, and exasperatingly, our stake on Maillot Blanc was halved due to him finishing in a dead heat for first. Still, we crack on.
TOP RATEDS AFTER DAY 10
393 RUNS
125 WINS
31.8% STRIKE RATE
41.29 PTS TO BSP
10.7% RETURN ON INVESTMENT
I am also recording the top rated horses finishing in the frame too, we've had 72 seconds and 51 thirds, so it's trapping the first two 50.1% of the time and the first three 63.1% of the time. That is pretty good, after 10 days. We just need a couple of bigger priced winners for lift off, we are due and of 37 that went off double digits to BSP we've had 5 second place ones. It just needs a couple of them to become winners now.
If we average all the 10 day periods of the backtesting, this is in the 43rd percentile for profit (49pts the middle), the 60th percentile for strike rate and the 38th percentile for return on investment (55pts the middle). So you can see how a couple of decent winners will make all the difference - they are almost certainly coming.
Note I'm taking the value bets off for now for time reasons and also because it's flagging anything the first 3 of the ratings no matter the number of runners. It needs adjusting so that it can only be a pick if the each way terms would justify it. I don't think it's right to make the 3rd ranked as a value pick if there are only 4 runners.
So today's picks attached and the Chester racecard. An 80.2% market correlation with morning lines.
AustinDillon75
Colt
The recovery continues - yesterday, we had 40 races and 14 winners and the BSP return was 9.29. Twirler at Windsor was our biggest winner at a BSP of 6.38. Again, it's feeling like a grind because only twice did we have back to back winners and we finished the day with a losing run of 5.
TOP RATEDS AFTER DAY 11
433 RUNS
139 WINS
32.0% STRIKE RATE
50.58 PTS TO BSP
11.7% RETURN ON INVESTMENT
The 49-60-38 percentiles from yesterday are now at 52 (profit) - 62 (strike rate) - 42 (ROI) for an 11 day period when compared to the backtesting. If it gets to 50-50-50 we know that we are aligned to the previous 4 years.
Note that I am working to a chosen 2% commission rate. The BSP doesn't drop hugely even if I were to adopt the 5%.
Attaching today's picks and the Chester card. Our correlation with the market for today sits at 77.1% so a bit lower than yesterday but the racing is overall very competitive.
TOP RATEDS AFTER DAY 11
433 RUNS
139 WINS
32.0% STRIKE RATE
50.58 PTS TO BSP
11.7% RETURN ON INVESTMENT
The 49-60-38 percentiles from yesterday are now at 52 (profit) - 62 (strike rate) - 42 (ROI) for an 11 day period when compared to the backtesting. If it gets to 50-50-50 we know that we are aligned to the previous 4 years.
Note that I am working to a chosen 2% commission rate. The BSP doesn't drop hugely even if I were to adopt the 5%.
Attaching today's picks and the Chester card. Our correlation with the market for today sits at 77.1% so a bit lower than yesterday but the racing is overall very competitive.
AustinDillon75
Colt
Will report on all that tomorrow but I know there are a lot of very good winners in there. I'm expecting the fourth consecutive profitable day.
AustinDillon75
Colt
So yesterday turned the profit we thought it would, 47 races and 15 winners and the BSP return was just 3.98. I was getting the feeling of being chinned a bit with a couple (10 top rated runners up) and Holloway Boy and Bradbury were two who certainly managed to snatch defeat from the jaws of victory. In the last 2 days we've had 4 beaten a short head, head, neck and neck respectively. We've had 1 winner by a neck but every other winner by a half length of more. Median margin of winners is 2.25 lengths and beaten losers 1.5 lengths.
The model is designed to grind things out a bit but it is somewhat becoming overdue the big price selection that pays out. We still only have 1 double figures winner in 12 days. The 12 day percentiles (profit-strike rate-ROI) are now 46-65-40.
Looking at backtesting these are the 12 day percentile stats:
.
So if we were in the 80th percentile, we'd have 104.81 pts profit over 12 days, a 33% strike rate and a 24.2% return on investment. So it's holding up and due that big day.
TOP RATEDS AFTER DAY 12
480 RUNS
154 WINS
32.1% STRIKE RATE
54.56 PTS TO BSP
11.4% RETURN ON INVESTMENT
The model is designed to grind things out a bit but it is somewhat becoming overdue the big price selection that pays out. We still only have 1 double figures winner in 12 days. The 12 day percentiles (profit-strike rate-ROI) are now 46-65-40.
Looking at backtesting these are the 12 day percentile stats:
.So if we were in the 80th percentile, we'd have 104.81 pts profit over 12 days, a 33% strike rate and a 24.2% return on investment. So it's holding up and due that big day.
TOP RATEDS AFTER DAY 12
480 RUNS
154 WINS
32.1% STRIKE RATE
54.56 PTS TO BSP
11.4% RETURN ON INVESTMENT
AustinDillon75
Colt
Today's rating, and the cards for both Ascot and Haydock. The cards can be subject to a lot of further development and insight, and I am thinking about putting in a simple checklist for horses too that gives red, amber and green to various characteristics and an at-a-glance insight into their prospects for the day.
AustinDillon75
Colt
Just to add, the correlation with the morning line today sits at 84%.
AustinDillon75
Colt
I am expecting a large chunk of the profit to be gone today. We'll report in the morning.
AustinDillon75
Colt
It sounds a bit like a blind defence of this but to be honest, we got 15 winners from 59 races but also had 14 2nds and 13 3rds. A lot of horses hitting the frame but a loss of 16.31 pts on the day. It's really annoying having had four profitable days in a row to give it almost all of it back and have little to show for that hard work.
TOP RATEDS AFTER DAY 12
539 RUNS
169 WINS
31.4% STRIKE RATE
38.25 PTS TO BSP
7.1% RETURN ON INVESTMENT
Today's picks soon.
TOP RATEDS AFTER DAY 12
539 RUNS
169 WINS
31.4% STRIKE RATE
38.25 PTS TO BSP
7.1% RETURN ON INVESTMENT
Today's picks soon.
AustinDillon75
Colt
Selections and card for Leopardstown. The market correlation today is at 86%.
Last edited:
AustinDillon75
Colt
A better day there, 22 bets and 8 wins (6 seconds) and a profit of 8.74 pts. Our biggest priced winner so far too, Moon Mission at Cork at 14.5 BSP.
TOP RATEDS AFTER DAY 13
562 RUNS
177 WINS
31.4% STRIKE RATE
45.98 PTS TO BSP
8.2% RETURN ON INVESTMENT
TOP RATEDS AFTER DAY 13
562 RUNS
177 WINS
31.4% STRIKE RATE
45.98 PTS TO BSP
8.2% RETURN ON INVESTMENT
AustinDillon75
Colt
Today's top rated selections. Some good action in Ireland later on today so attaching the two cards for those meetings. The correlation to the morning line for today's 6 meetings is 80.2%.
Please again note, the model does not pay any attention whatsoever to today's market. It forms it's own opinion on the races completely independent. In time we may decide to create our own tissue prices based on the ratings.
Please again note, the model does not pay any attention whatsoever to today's market. It forms it's own opinion on the races completely independent. In time we may decide to create our own tissue prices based on the ratings.
Last edited:
AustinDillon75
Colt
The recovery continues. 40 selections today, 14 winners (5 2nd, 5 3rd). A BSP profit of 11.94 pts. So we are moving in the right direction again - it still lacks those one or two larger priced winners and we are seeing the strong strike rate doing a lot of heavy lifting at present.
TOP RATEDS AFTER DAY 14
602 RUNS
191 WINS
31.7% STRIKE RATE
57.92 PTS TO BSP
9.6% RETURN ON INVESTMENT
The profit - strike rate - ROI percentiles to the 14 day backtesting are at 40-59-33. As I say, nothing a couple of large priced winners won't resolve. We've still only had 2 double figure winners from the 191. Back tomorrow.
TOP RATEDS AFTER DAY 14
602 RUNS
191 WINS
31.7% STRIKE RATE
57.92 PTS TO BSP
9.6% RETURN ON INVESTMENT
The profit - strike rate - ROI percentiles to the 14 day backtesting are at 40-59-33. As I say, nothing a couple of large priced winners won't resolve. We've still only had 2 double figure winners from the 191. Back tomorrow.
AustinDillon75
Colt
Today's ratings. No cards attached today. If we can just get through the next 3 or 4 days in profit or at least with no big loss we will likely be staying on track. It is designed with the strike rate to make lots of small gains and it's done that but it's not quite had that big winner that would cover all the losers.
I'm attaching a list of all the bets so far and the BSPs. Profit graph with bets on the horizontal and profit in units on the vertical.
I'm attaching a list of all the bets so far and the BSPs. Profit graph with bets on the horizontal and profit in units on the vertical.
Sullybomb
Gelding
AustinDillon75 That is a lot of bets per day (average 43) are you automating all of these or manually betting? Or still in the trial period and not actually betting for real £££?I tried to do manual betting when I was running my own and it was too time consuming, I have managed to automate it all now and can monitor any issues on my phone through alerts, far easier. Something to consider if/when you go live.
Cheers
AustinDillon75
Colt
Thanks. No, I have backtested for 4 years and want to make sure that real forward testing is delivering the same outcomes without data leakage. I am not putting any real money on this for a good while to come.
I tried to do manual betting when I was running my own and it was too time consuming, I have managed to automate it all now and can monitor any issues on my phone through alerts, far easier. Something to consider if/when you go live.
Cheers
It is obviously a lot of bets but if it replicates the backtesting I will go live with a modest bank and see if it scales up in time. It is as important me to have a nice rating set too and the backtesting does actually point to those value bets delivering lower strike but bigger roi so I am definitely coming back to them.
I will do another sheet later to see where we would be if we reset the stake daily to 1% of the bank. I think an original bank would likely have doubled by now.
Sullybomb
Gelding
I started with £500 bank and it's now over £3000 so it's absolutely possible. I've had just shy of 1000 bets since the 16th Jan when I went live and the LLR has been 26 which was when the changeover from AW to Turf started.
Cheers
AustinDillon75
Colt
I have to be honest and say that was a pretty infuriating day of two halves and we ended with a losing run of 9 races. 38 bets, 16 winners (9 2nd, 4 3rd). Accepting that there were some 2 and 3 runner races today distorting slightly.
A BSP profit of 5.33pts. We are really just grinding this out at present more so than I'd have hoped but sticking it thoroughly, and it's the 7th day of the last 8 where a profit has been delivered.
TOP RATEDS AFTER DAY 15
639 RUNS
207 WINS
32.4% STRIKE RATE
64.25 PTS TO BSP
10.1% RETURN ON INVESTMENT
The profit - strike rate - ROI percentiles to the 14 day backtesting are at 43-72-34. Once all three are at 50 it will demonstrate clear equilibrium between our backtesting and the daily output.
A BSP profit of 5.33pts. We are really just grinding this out at present more so than I'd have hoped but sticking it thoroughly, and it's the 7th day of the last 8 where a profit has been delivered.
TOP RATEDS AFTER DAY 15
639 RUNS
207 WINS
32.4% STRIKE RATE
64.25 PTS TO BSP
10.1% RETURN ON INVESTMENT
The profit - strike rate - ROI percentiles to the 14 day backtesting are at 43-72-34. Once all three are at 50 it will demonstrate clear equilibrium between our backtesting and the daily output.
AustinDillon75
Colt
Just to pick up on that 1% daily bankroll, if we'd adjusted for a 1% of the £200 bank stake daily, we'd currently be on a bank of £297.60, so we'd be staking £2.98 on each runner tomorrow.
I've also decided to have another look at the value bets and do an assessment of them - only counting those where place terms would apply, and identifying the BSP to win and place. I'm going to retrieve them and start adding them to the book because I am sure this might well point to profits that are more volatile but could require fewer bets.
I've also decided to have another look at the value bets and do an assessment of them - only counting those where place terms would apply, and identifying the BSP to win and place. I'm going to retrieve them and start adding them to the book because I am sure this might well point to profits that are more volatile but could require fewer bets.