Gemini is ahead of the other chatbots, at -14.6%.
The full order after 91 days and 37,288 settled bets, best to worst:
| Tipster | Settled bets | Return |
|---|---|---|
| Gemini | 5,793 | -14.6% |
| Grok | 5,793 | -15.2% |
| Claude | 5,696 | -16.1% |
| ChatGPT | 5,798 | -18.0% |
| DeepSeek | 5,567 | -18.8% |
Every one of them is losing money. So is our own model, and so is a plain baseline that simply backs the favourite in every race.
That last comparison is the one worth sitting with.
The 5 publicly available chatbots in the league are given the same racecard every day and asked for their selections. The picks are logged before the races run, settled at starting price, and published whether they win or lose. Nothing is retro-fitted and nothing is quietly dropped.
Each model runs two arms. The blind arm sees the runners and the form but not the prices. The informed arm sees the market as well. The difference between them tells you how much a model is simply reading the odds back to us.
For Gemini, blind returns -12.1% and informed -17.4%. The size of that gap, and its direction, are not uniform across the models, and that is one of the more interesting things the league has produced.
We publish the whole thing at the Silicon Tipster League, including every daily pick.
People ask what this costs to run. The tipsters are the cheap part of it. The betting is what costs money.
Not intelligence, as far as we can tell. What separates them is how close each one sticks to the favourite.
Gemini backs the favourite 12% of the time and leads. The model that backs favourites most often is not the leader, and the one that backs them least is not last. The relationship is real but loose.
The more useful reading is that all five sit within a few points of each other and of the favourite baseline, which suggests they are all doing something fairly similar: reading the form, arriving near the market's opinion, and then paying the charge like everyone else.
We should be careful about the order itself. 91 days is not a long time, the gap between first and last is a few percentage points, and these standings have moved before. We wrote a whole piece on exactly that, why the standings keep moving, because a lead this size is not yet evidence of skill.
None of these pages is a tip, and none describes a way to win. They describe what betting costs, which is a different and more reliable subject. If your betting has stopped being fun, BeGambleAware has free, confidential help, and the National Gambling Helpline is on 0808 8020 133.
On our league, Gemini leads at -14.6% over 5,793 settled bets. It is losing money, as are the other four.
They are published daily and settled in public, and every one of them is currently unprofitable. We publish them as an experiment, not as a tipping service.
At starting price, level stakes, with no best odds guaranteed and no commission. Voids are excluded.
The blind arm does not see the betting odds. The informed arm does. Comparing them shows how much each model is simply reflecting the market back.
No. The spread between first and last is small relative to the noise in a sample this size, and the order has changed more than once already.
Every figure here is pulled live from our data and nothing beats the bookmaker's margin. For whether anyone holds a real edge, see our track record. 18+, please bet responsibly.