The evidence
Every tipping service shows you a profit curve. Almost none will tell you how it was calculated, over what sample, or whether it survives a proper test. This page is our answer to all three — including the parts that do not flatter us.
Not a highlights reel. Every Sharp selection since the product began, counted in the month it ran, with the breakdown underneath. Each one was published before its race ran, at the odds shown, and settled from the official result. Until September 2026 they went out inside the wider Signals group, and some of those days are still being restored to this site.
No month has been taken out. A losing month appears in the table below in red, on this page, without comment from us.
Sharp — the record
42.1%
Strike rate
+19.1%
Return
Average winning price $2.83. Every month Sharp has run is in the table below, in order, with its own count.
Profitable months
5 of 7
Longest losing run
9 bets
Worst drawdown
9.0 stakes
Excluding Heavy going
51 bets · 45.1% strike · +27.8% return (+14.2 stakes)
We no longer bet Heavy tracks. 6 Sharp bets went out before that rule was in place and they are still counted in the headline figure above — we do not remove published bets after the fact. This second line is the record over the bets we would take today. Both are real; the gap between them is what those 6 cost.
Every month, flat stakes
Not a selected window. Every month Sharp has run, in order. A losing month appears here in red. There are 2 below.
| Month | Bets | Won | Return |
|---|---|---|---|
| Apr 2026 | 1 | 1 | +200.0% |
| May 2026 | 5 | 3 | +70.4% |
| Jun 2026 | 11 | 4 | +2.0% |
| Jul 2026 | 6 | 2 | -10.7% |
| Aug 2026 | 12 | 7 | +67.2% |
| Sep 2026 | 9 | 4 | +29.0% |
| Oct 2026 | 13 | 3 | -37.7% |
The table above is that record, month by month — when a month goes backwards it appears here in red. It runs to 57 settled bets at a 42.1% strike rate, average price $2.83, each one published before its race and settled from the official result. 57 bets is a small record. A run this good can still be luck, and we can't yet show that it isn't. We publish it either way.
Use this on us. Then use it on whoever you are paying now.
How many bets is that figure based on?
Under a few hundred, a strike rate tells you very little and a return figure tells you almost nothing. Ask for the count, not just the percentage. Ours is on this page — the total under the table above, and every month's count in it.
Were the selections published before the races ran?
A record assembled afterwards is not a record. Ours are timestamped at publication and locked — the archive shows what was sent, when.
Are the losing runs still in there?
The most common trick is quietly dropping a bad month, or a losing tier. We drop neither. If a stretch went badly, you can find it.
At what price was the return calculated?
Returns quoted at starting price, or at the best price available anywhere, are not the price you got. We record the odds visible at the moment we published.
Has the rule been changed to fit the results?
A filter re-tuned after a bad run will always look good in hindsight. Sharp's thresholds are frozen; changing them means publishing a new version, not quietly editing the old one.
Not a guaranteed return. Nobody can honestly sell you one, and the operators who imply it are the reason this industry has the reputation it does. What you get is narrower and, we think, more useful:
A disciplined card, every race day
Three to six selections chosen the same way every single time, with no mood, no hunches and no chasing yesterday's loss.
Stakes that won't ruin you
Flat staking, because on our own numbers the aggressive alternative produced a deeper drawdown for a worse return. The plan is boring on purpose.
The time back
The model reads every runner at every qualifying meeting before breakfast. That is the part you are really buying.
A record you can audit
Complete, timestamped, settled by machine, and published whether it is going well or not.
If a tipping service is losing you four to ten percent of turnover — which, on the published numbers, plenty are — then a disciplined process with an honest ledger is a real improvement. That is the claim. We are not going to dress it up as anything more until the evidence says we can.