Every morning, thousands of punters across the UK ask the same question: which horses should I back today? Most tipsters answer that question with experience, intuition, and a quick scan of the Racing Post. At TheUltimateTipster, we answer it with artificial intelligence that analyses over 150 data points for every single runner on today's card. In this article, we'll pull back the curtain and show you exactly how our AI generates today's racing tips — from raw data ingestion to the final selections that land in your members' area each morning.
Understanding this process matters because it explains why AI-generated tips are fundamentally different from human tips. A human tipster, no matter how experienced, can only process a fraction of the available data. They're subject to biases, fatigue, and the temptation to follow the crowd. Our AI has none of these limitations. It processes every runner with the same rigour, whether it's the 2:15 at Catterick or the feature race at Ascot.
The Morning Data Pipeline
Our system starts working in the early hours of the morning, long before most punters have opened their eyes. The first step is ingesting the day's race cards — every declared runner at every UK and Irish meeting. This includes basic information like the horse's name, trainer, jockey, draw position, and official rating, but that's just the starting point.
The AI then pulls in a vast array of supplementary data for every runner. This includes the horse's complete career form — not just the last few runs that a human might glance at, but every single run the horse has ever had. It analyses going preferences across the horse's entire career, identifying whether the horse performs better on firm ground, soft ground, or heavy going. It examines distance performance, looking at whether the horse's best runs have come over five furlongs, a mile, or two miles.
Trainer and jockey statistics are pulled in with granular detail. Rather than simply noting that a trainer has a 15% strike rate, the AI examines the trainer's performance at this specific course, over this specific distance, in this class of race, and during this time of year. These conditional statistics often reveal patterns that global averages completely miss. A trainer might have a modest overall win rate but an exceptional record when sending horses to a specific track.
Speed Figures and Form Analysis
One of the most powerful inputs to our model is speed figure data. Speed figures represent how fast a horse actually ran in each of its previous races, adjusted for going conditions, wind, and the overall pace of the race. This is far more informative than simple finishing positions because it accounts for the quality of the race.
For example, a horse that finished third in a Group 1 race might have produced a speed figure of 115, whilst a horse that won a Class 5 handicap might have produced a speed figure of only 85. The finishing positions alone would suggest the Class 5 winner is in better form, but the speed figures reveal that the Group 1 third is a far superior performer. Our AI uses these figures to compare horses on a level playing field.
The form analysis goes deeper than most punters realise. The AI examines form cycles — the pattern of a horse's performances over time. Is the horse improving, declining, or maintaining a consistent level? Has it just returned from a break, and if so, how does its trainer's record with fresh horses compare to the average? Has the horse been running consistently without a break, suggesting it might be due a rest or a below-par effort?
Market Data and Smart Money
Perhaps the most dynamic input is live market data. Before the first race of the day, our system begins monitoring betting market movements. When significant money flows onto a particular horse — causing its odds to shorten by 30% or more — the AI flags this as a potential market mover.
Market movements matter because they often represent informed money. When connections to a stable back their horse heavily, or when professional betting syndicates identify value, the market moves. Our AI doesn't just note that a horse has shortened in price; it analyses the timing, magnitude, and pattern of the movement to assess whether it represents genuine "smart money" or simply public sentiment following a newspaper tip.
This market intelligence is combined with our form-based analysis. When a horse that our model already rates highly also attracts significant market support, the convergence of signals creates a particularly strong selection. These convergence picks have historically produced our best results.
The Scoring System
Once all data has been ingested and processed, the AI assigns a composite score to every runner. This score is built from multiple weighted components:
Form component: Based on recent speed figures, finishing positions, and form trajectory. A horse showing improving form scores higher than one in decline.
Going suitability: How well does today's going match the horse's proven preferences? A horse with a 40% win rate on soft ground running on soft ground scores much higher than one with a 5% win rate on the same surface.
Course and distance: Has the horse performed well at this specific course? Over this distance? The AI identifies course specialists that human tipsters might overlook.
Trainer and jockey: Current form of both trainer and jockey, their record at this course, and their record as a combination. A trainer and jockey who have a 25% win rate together at this course deserve attention.
Class assessment: Is the horse well-handicapped? Is it dropping in class from a higher level, suggesting it might be ahead of today's rivals? Or is it stepping up in class, which statistically reduces win probability?
Draw bias: On certain courses and over certain distances, the draw position has a statistically significant impact on results. Our AI incorporates draw bias data calculated from thousands of historical results at each course.
Each component is weighted according to its predictive power, and the final composite score determines whether a horse qualifies as a selection and which tier it falls into.
The Tier System
Not all selections are created equal, and our tier system reflects this. The AI assigns each qualifying horse to a tier based on its composite score and the nature of its edge:
Banker: The highest-confidence selections. These are horses where multiple strong signals converge — excellent form, ideal conditions, strong market support, and a clear class edge. Bankers are designed for punters who want confident, higher-stake selections.
5-Star: Elite contrarian picks where the AI has identified a significant edge that the wider market may have undervalued. These selections often offer better odds than Bankers because the market hasn't fully recognised their chance.
Hidden Gem: Each-way value at bigger prices, typically 12/1 or higher. These selections are specifically identified for each-way betting, where the place element provides downside protection whilst the win element offers a significant return.
4-Star and Extra: Solid selections with a clear positive edge, suitable for level-stakes betting. These represent genuine value without the extreme confidence of the top tiers.
This tiered approach allows punters to adjust their staking according to their risk appetite and bankroll management strategy. You can check how each tier performs on our Recent Winners page.
Why AI Beats Gut Feeling
The fundamental advantage of AI over human tipsters is consistency and completeness. A human tipster might study 30 or 40 horses in detail on a given day. Our AI analyses every single runner — often 100 or more — with the same depth and rigour. It never has an off day, never gets tired, and never falls for a horse just because it has an attractive name or an exciting running style.
Human tipsters are also subject to cognitive biases. Recency bias means they overweight a horse's most recent run. Confirmation bias means they look for evidence that supports their initial impression. Anchoring bias means they're influenced by a horse's odds rather than its objective chance. The AI is immune to all of these.
That said, our AI isn't magic. It can't predict the unpredictable — a horse might stumble at the start, get boxed in, or simply have an off day. What it can do is identify, day after day, the runners that have the best chance of winning at odds that represent genuine value. Over hundreds of selections, that edge compounds into meaningful profit.
Seeing It in Action
The best way to understand how our AI picks today's tips is to try it yourself. Every selection comes with a confidence score and clear reasoning, so you can see exactly why the AI has chosen each horse. You won't find vague statements like "looks in good form" — you'll see specific data points and analysis.
Visit our Recent Winners page to see the full, transparent record of every selection and every outcome. We publish everything — wins, places, and losses — because we believe that's the only way to build genuine trust.
Ready to see how AI picks today's racing tips? Start your free 14-day trial and get instant access to today's selections, market movers, and raiders. No payment required upfront — just data-driven tips you can verify for yourself.
Frequently Asked Questions
Q: How does your AI horse racing predictor work?
A: Our AI ingests over 150 data points per runner — form, speed figures, going preferences, trainer/jockey records, draw bias, market signals, sectional times, breeding angles and more. Three independent reasoning engines (Agentic GENIUS, AI Form Reader and Specky) each score the race separately, then a moderator agent debates the picks to surface high-conviction selections.
Q: Which AI model do you use to generate tips?
A: We use a mix of Google Gemini 2.5 Pro for deep qualitative form reading, Gemini 2.5 Flash for the Golden Zone reasoning layer, and Moonshot Kimi K2 for contrarian value analysis. Each model is given a different role, prompt and data window so they reason independently and we get genuine diversity of opinion rather than three flavours of the same answer.
Q: What time does the AI publish today's racing tips?
A: The full daily tips run completes by 10:30 UK time. Selections are time-stamped, written to a public database, and emailed to subscribers immediately afterwards. There is no editorial pause — tips go live the moment the AI finalises them.
Q: Can AI really beat the betting markets long-term?
A: The markets are extremely efficient at the top of the card, but at prices of 3/1 and bigger — where there is less liquidity and more public money chasing names — there is consistent inefficiency a well-built AI can exploit. Our 90-day rolling ROI across endorsed tiers has stayed positive even during the model's worst weeks.
Q: How is your AI different from At The Races or Sporting Life predictors?
A: Most of those tools publish a single power rating per horse and stop. Ours runs a multi-engine debate, applies a strict 3.0 minimum-price value filter, excludes race types where the data is too thin to call (maidens, sellers, junior NH), and self-learns from its own results overnight via the Nightly Debrief loop.
Q: Do you publish losers as well as winners?
A: Yes — every selection ever published is recorded with its result. Our Recent Winners page shows full month-by-month strike rate, ROI and placed stats including the losing days. There is no cherry-picking and no quiet deletion of bad weeks.