The same disagreement test as above, run specifically on Flat handicaps -- the segment with the most historical validation, now being tested forward on live races.
Settlement note: from 8 October 2026 every bet on this page is a straight win bet, because the strategy backs the model's top pick to win, and all earlier bets have been re-settled the same way so the record is consistent. Before that date, picks priced above 4/1 were settled each-way.
-61.97%ROI
7.7%Strike rate
-24.17Total profit
39.0Total staked
393W 0P 36L · 1 void
0Open bets
Closing line value
Our price compared with
Bets
Median CLV
Average CLV
Beat price
Bookmaker SP, same-day prices (from 6 Oct 2026)
10
+0.0%
+2.2%
40%
Bookmaker SP, all bets
39
+4.5%
-2.6%
51%
Betfair SP, same-day prices (from 6 Oct 2026)
10
-16.0%
-12.2%
10%
Betfair SP, all bets
38
-18.3%
-18.1%
21%
Betfair SP vs the typical runner at the same price, same-day
10
+0.9%
+6.6%
60%
Compares the price we logged with the price the market closed at: a positive figure means we took a better price than the market closed at. It is an early sign of whether the model finds real value, well before profit and loss settles down, but small samples swing a lot. Betfair SP is a sharper benchmark than the bookmaker starting price (bookmaker prices carry a margin, especially on longer-priced horses) and is shown before Betfair commission, so every runner normally looks negative against it and the gap widens with the price. The "typical runner" row corrects for that: it compares each bet with the average runner in the same price band (under 4, 4 to 8, 8 to 15, 15 plus) since 6 October, in percentage points, so a positive figure means the model's picks closed better than a typical horse at that price. Bets logged before 6 October 2026 could be logged the evening before, on thinner early prices, so they are shown separately.
Drawdowns, losing runs and monthsthe bad spells as well as the good, from 39 settled bets
Worked out from the bets as logged (level stakes, void bets left out), in race order. A drawdown is the biggest fall in running profit from a previous high point. An each-way bet that placed but still lost money counts as a losing bet. Losing runs this long or longer can happen again, whatever the results so far.
Worst drawdown
−24.17 units
24 Sep 2026 to 15 Oct 2026 · not yet back to that high
Below its best
−24.17 units
best running profit so far: +0.00 units
Longest run without a winner
18 bets
02 Oct to 15 Oct 2026 · current run: 18
Longest run of losing bets
18 bets
02 Oct to 15 Oct 2026 · current run: 18
Month
Bets
Wins
Staked
Profit
ROI
Running profit
Oct 2026
22
2
22.00
-11.67
-53.0%
-24.17
Sep 2026
17
1
17.00
-12.50
-73.5%
-12.50
Weekly resultsprofit and ROI for each week (Monday to Sunday), by race date
Each row is one week of settled bets. ROI is profit divided by units staked that week. A single week is a small sample and swings a lot, so look at the cumulative ROI column for the longer picture. The current week is still in progress. Void bets are not counted.
Week
Bets
Wins
Staked
Profit
ROI
Cumulative ROI
12 Oct – 18 Oct
4
0
4.00
-4.00
-100.0%
-62.0%
05 Oct – 11 Oct
10
0
10.00
-10.00
-100.0%
-57.6%
28 Sep – 04 Oct
17
3
17.00
-2.17
-12.8%
-40.7%
21 Sep – 27 Sep
8
0
8.00
-8.00
-100.0%
-100.0%
Staking simulatorflat staking against Kelly staking on this strategy's 39 settled bets
This replays the settled bets in race order starting with a bankroll of 100 units, using the results exactly as they were settled. Flat staking puts 1 unit on every bet. Kelly staking puts a share of the current bankroll on each bet, sized by how much value the model saw in it (expected value ÷ (odds − 1)), so bigger edges get bigger stakes and the stakes grow and shrink with the bankroll. Quarter and half Kelly use a quarter or half of that share, which is how Kelly is usually used because full Kelly swings a lot. No single bet is capped above 10% of the bankroll. Bets the model rated at zero or negative value get no Kelly stake: 26 of 39 bets here had positive expected value. Each-way bets use the same formula as an approximation, and bets that overlapped on the same day are treated as one after another. With a small number of bets, the result mostly shows how lucky or unlucky the run was, not which staking plan is better.
Staking plan
Bets staked
Final bankroll
Change
Biggest drop
Flat staking (1 unit a bet)
39
75.8
-24.2%
-24%
Quarter Kelly
26
98.5
-1.5%
-8%
Half Kelly
26
95.8
-4.2%
-16%
Full Kelly
26
87.9
-12.1%
-30%
Why Kelly can lose more than flat staking. Kelly stakes more when the model sees more value. When the model's edge is overstated, the biggest stakes go on the bets where the model is most wrong, and losses compound: a bankroll that falls by half has to double just to get back to where it started. That is why we judge every method at flat stakes.