The AI That Led the Tipster League Had Its Worst Month

In late August, Gemini sat top of the Silicon Tipster League on percentage return; its combined record now stands at -11.4% from 3,888 settled picks over 62 days. Read as a single number, a league lead looks like a model pulling clear of the field. Read against its own month, it looked like something closer to the opposite.

On 24 July, Gemini's running total in the league was comfortably in profit, its best day of the run to that point. By 16 August it had fallen to its worst point of the run, deep in the red. The Ebor Festival week clawed back a real chunk of that through to the Saturday, before the line drifted down again through the rest of August. Today it stands at −£444. That is not a model climbing a hill. It is a model that peaked in late July and has spent most of the time since digging out of a hole, currently still in it.

None of that is a criticism specific to Gemini. Every entrant in the league is losing to the book's own margin, and the honest read of any single model's month is that a few results either side would have told a different story. What makes this one worth writing up is what it does to the meaning of "leading the table": at the end of August, Gemini's position was the best percentage return among the eight entrants, and its own August had, in cash terms, been one of the roughest stretches anyone here has had. Both things were true at once, and the table on its own only shows you the first one.

Here is the number that turns this from a losing streak into something worth reading: The Favourite, our baseline entrant that does no analysis at all and simply backs the shortest price in every race, currently shows a smaller cash loss than Gemini does, at −£231 against Gemini's −£444. By pounds and pence, backing the market's own top pick every time has cost less than the model that sat top of the leaderboard in late August.

When this was written, Gemini's percentage return was the better of the two (the live figures on the board now read -11.4% against The Favourite's -11.3%), because ROI and cash profit answer two different questions. Return on investment asks how much came back for every pound risked. The Favourite has settled 2,035 picks to Gemini's 3,888, a smaller book on which a smaller loss is easier to hold. Gemini has risked close to twice as much in total stake, so a larger cash loss on a larger book can still be the better rate of loss per pound. ROI is the fairer way to compare two records built on different numbers of bets, which is exactly why it is the number the league is ranked on.

The reason this is worth spelling out rather than just quoting the leaderboard is that either number, read alone, tells a flattering half-story. At the end of August the percentage said Gemini was ahead, while the cash said the simplest possible strategy on the board had bled less that month than the model in first place. Both were correct, and a reader who only saw one of them would walk away with a cleaner story than the numbers actually support.

What a lead this size is worth

62 days and 24,675 settled picks is a bigger sample than the two or three weeks this league started with, and it is still not enough to call a winner. The gap at the top of the table in percentage terms stays small enough that a single good or bad weekend can close it or widen it again, the way Gemini's own month swung from comfortably ahead of break-even to its worst point of the year and partway back.

Backing every favourite in British and Irish racing loses money against the book's own margin, which is exactly what it is supposed to do here: it is the bar every tipster in the league has to clear, not a strategy we are recommending. That it held up better in cash terms than the model leading the table that month is a fact about one month, not a verdict on either approach.

The standing rule for reading anything on this board still applies: no entrant here has beaten the margin built into the price of every bet, and a model leading today has not solved anything the rest of the field has not also failed to solve. Follow the current standings on the Silicon Tipster League, read why the standings keep moving, or see what changes when an AI sees the odds. Nothing on that board is a tip, and none of it is betting advice. 18+, please gamble responsibly.