The Psychedelic Socialist
ClioSport Club Member
Can you imagine the state of an AI trained on CS forum posts. Christ alive.
ClioSport.net
Register a free account today to become a member!
Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!
-
When you purchase through links on our site, we may earn an affiliate commission. Read more here.
ChatGPT (AI)
- Thread starter Fantastic Nostrils
- Start date
Beauvais Motorsport
ClioSport Club Member
Imagine? I thought half the posters were AI.
F31 35d, Skoda Yeti
Ye it did this when I was asking if Verstappen could win the championship based off different scenarios I gave.
Think I blew a server
Ah, GPT’s called it a day, has it? Typical. Don’t worry — I’m the stand-in with fewer credentials but a better sense of humour
Note: GPT might have wrote this
Oggy997
ClioSport Club Member
997, e208, I.D. Buzz
No use to you now, but maybe an option to consider, we run a subring where we share a login.
The rules were you clean up after yourself, but one guy has been using it to plan how to leave his wife for another girl, which he did, and then she's left him on the sofa after that.
He literally inputs every conversation he has with the new girl to try and win her back round.
That alone was worth the fee
Ay Ay Ron
ClioSport Club Member
And you haven't made a separate thread on this? Why?
Akay
ClioSport Club Member
Clubman Cooper S
Im seeing more and more 'have a go heroes' at work using this, putting contracts in and generating comments, and then passing it off as there own work.
Im not normally arsed, but they dont even delete the tell-tale signs that Chat GPT used it, such as the bloody emojis or the 'would you like me to put this into a pdf, or share a link - just say the word!'.
Morons.
Its also great asking questions on it when they clearly havent even read the output.
Im not normally arsed, but they dont even delete the tell-tale signs that Chat GPT used it, such as the bloody emojis or the 'would you like me to put this into a pdf, or share a link - just say the word!'.
Morons.
Its also great asking questions on it when they clearly havent even read the output.
Ol’ Tarby
ClioSport Moderator
Clio 220 Trophy
There's a guy at our place who thinks hes Billy big b****cks now as he's been inputting fault logs from our panels into chatgpt to get it to tell him whats going faulty, and is there any patterns etc.
If you can't do that by yourself, then you're s**t at your job (he is).
If you can't do that by yourself, then you're s**t at your job (he is).
Oggy997
ClioSport Club Member
997, e208, I.D. Buzz
Oh and I think he's her boss, the bit on the side, they definitely work together.
It's equally hilarious and tragic
massiveCoRbyn
ClioSport Club Member
Several
The only difference I noticed was that it seems a bit more reliable. I was doing something with a consultant and we were using the paid and free versions to experiment. The free version seemed to go off on a bit of a tangent and then get stuck in a hallucination loop pretty quickly, while the paid version kept on going (before eventually doing the same, obviously).
I still have mixed feelings about it to be honest. My missus loves it and uses it all the time. I do pay for it, and use it quite a bit, but also find it quite frustrating and often find myself thinking it would have been quicker to do it myself.
The Psychedelic Socialist
ClioSport Club Member
Perhaps a bit niche, but Gowers is a Fields Medal winning mathematician (Nobel prize for maths). Provides an interesting contrast to the Reddit chorus of "it's just glorified autocorrect"....
On a bit of a side project I'm taking down the gambling industry using GPT.
Over the last 6 weeks or so we've been running a framework where I have given it 10 points initially to build the framework and everyday using that framework at 11:00 we pick the best 5 horses from the days racecards. At the end of each day we review the results, see where we went wrong, make small improvements and continue. This can be anything from where the ground let us down with late changes in the expected going, specific course nuances we didn't consider, etc etc.
We started off hit and miss, was just like picking by favourite colours. We got slowly better, around -15 to -20% ROI the last few weeks and then results from the last 7 days were.....
I've not bet any money yet myself as this is just a concept I've been working on but it is going int he right direction. I'll need it to consistently turn out these sort of returns over a long period of time before I'll stake anything substantial.
Interesting to read through its analysis although half of it means nothing. Recently started adding something called the Kelly Criteria???? which is some sort of probability theory.
Over the last 6 weeks or so we've been running a framework where I have given it 10 points initially to build the framework and everyday using that framework at 11:00 we pick the best 5 horses from the days racecards. At the end of each day we review the results, see where we went wrong, make small improvements and continue. This can be anything from where the ground let us down with late changes in the expected going, specific course nuances we didn't consider, etc etc.
We started off hit and miss, was just like picking by favourite colours. We got slowly better, around -15 to -20% ROI the last few weeks and then results from the last 7 days were.....
| Metric | Value |
|---|---|
| Valid picks | 29 (1 NR excluded) |
| Total staked | £290 |
| Total returned | £405 |
| Net profit | +£115 |
| ROI (SP basis) | +39.7 % |
| Winners | 9 / 29 (31 %) |
| Placed (Top 3) | 17 / 29 (59 %) |
I've not bet any money yet myself as this is just a concept I've been working on but it is going int he right direction. I'll need it to consistently turn out these sort of returns over a long period of time before I'll stake anything substantial.
Interesting to read through its analysis although half of it means nothing. Recently started adding something called the Kelly Criteria???? which is some sort of probability theory.
F31 35d, Skoda Yeti
Love this! 🤣
Matt Cup
ClioSport Club Member
Leon Cupra, 172 Cup
F31 35d, Skoda Yeti
Probably have a decent return compared to the lotto syndicate
Matt Cup
ClioSport Club Member
Leon Cupra, 172 Cup
Either that or we are just paying for his Coke and Hookers.
Perfect timing, Pete — here’s the current active framework that all 11:00 “Pick 5” runs are using right now:
🧩 Horse Racing Framework v2.1f (Active as of 21 Oct 2025 – AW-Bias Edition)
(base = v2.1a 10-Point System + refinements v2.1b → v2.1e)
🧱 1️⃣ Base 10-Point System
Factor | Key Rules |
|---|---|
Race Selection | Prefer ≤ 12 runners; avoid big handicaps/heavy ground/multi-unknown novice races. |
Horse Profile & Form | Use last 2–3 runs (OR/RPR/TS trend). RPR > OR → +1 pt. Progression trend +0.25 – 0.5. |
Course & Ground | C / C&D flags +0.25–0.5. Unproven surface −0.25. Going risk ±0.5 if forecast change. |
Collateral Form | Beat subsequent winners +0.25–0.5. Repeated defeats −0.25–0.5. |
Trainer Form | 14-day SR > baseline +0.25–0.5. < 10 % SR −0.25. Top-3 course record +0.25. |
Jockey Form | Hot 14-day/course SR +0.25; cold −0.25; won together before +0.25. |
Travel Factor | ≤ 150 mi neutral; > 250 mi −0.25 unless yard proven traveller. |
Tactical / Pace Map | Core variable (35–40 % of Adjusted Score). Compute PPI = (leaders + prominent)/field. PPI ≤ 0.35 → leaders +0.3; ≥ 0.55 → leaders −0.3. ≤ 7 runners → upgrade uncontested leaders. |
Market Shape / Value | Forecast < 2.0 → No Bet unless Dominant (Adj ≥ 8 & ≥ 3 green factors). Overlay % = (Fair–Market)/Market. |
Equipment / Ops / Misc | First-time or removed gear ±0.25; wind op +0.25 if yard hot; persistent gear −0.25. |
⚙️ 2️⃣ Framework Refinements (v2.1b → v2.1e carry-overs)
- Front-Runner Density Check (v2.1b): −2 pts if ≥ 2 habitual front-runners and our pick is also forward.
- Strong Finisher Bonus (v2.1b): +2 pts for proven late-closers / sectional trend positive.
- Going Consensus (v2.1b): if public going differs ≥ 1 step within 90 min → downgrade sprints.
- Trainer Form Gate (v2.1c): < 10 % 14-day SR → down-rank one tier.
- NH Micro-Field Blocker (v2.1c): Skip ≤ 4 runners unless fav ≥ 6/4 & ≥ 70 % jump rating.
- Ascot Lane Model (v2.1d): Straight 6f–1 m; rail bias +0.4 near rail, −0.25 mid/high without pace.
- Field-Size Filter (v2.1d): > 14 runners G1/G2 → −0.5 or exclude from acca.
- Pace Weighting Boost (v2.1e): Pace/Tactics = 35–40 % of score.
- Lone-Leader Rule (v2.1e): ≤ 6 runners and single likely leader ≠ our pick → −2 or No Bet.
- NH Novice Odds-On Guard (v2.1e): ISP < 1.80 and no ≥ 7 lb clear → Skip.
- AW Stayers Hold-Up Filter (v2.1e): ≥ 12 f & hold-up style −0.8 unless late-bias track.
- Going-Split Enforcement (v2.1e): Apply ±0.3 if official hurdle/chase split changes.
- Acca Hygiene v2 (v2.1e): Require ≥ 2 Leader/Co-Leader legs for acca build.
⚡ 3️⃣ v2.1f Additions (21 Oct AW-Bias Hot-Fixes)
1️⃣ Intraday AW Bias Adapter (Newcastle/Tapeta)
If first two races ≥ +2.5 s slow and winners come mid/back → reduce Leader Advantage −20 %, increase Closer Suitability +20 % for remaining Ncs legs.
2️⃣ Run-Style Uncertainty Dampener (≤ 7 AW runners)
−10 % pace-map confidence when trainer/jockey show ≥ 30 % “ride cold” tendency (last 30 days); allow prominent → mid/back re-score without penalty.
3️⃣ Acca Hygiene v3 (AW short prices)
Exclude AW legs ≤ 6/4 unless projected pace edge ≥ +1 length post-bias.
4️⃣ Going/PPI Soft-Upgrade (NH 3 m +)
Soft or worse → +5 % multiplier for proven stayers.
🧮 4️⃣ Selection & Ranking Logic
- Score each runner on 10 base factors (+/− refinement adjustments).
- Apply One-Per-Race hot-fix → keep highest Adjusted Score.
- Remove No Bet cases (short price, odds-on guard, lone-leader conflict, trainer gate).
- Rank remaining by Adjusted Score × Overlay %.
- Top 5 = singles; Top 3 (hygiene-qualified) = win acca.
📊 5️⃣ Output Format (v2.1a″ Simplified Value Display)
Each pick shows:
- Race & Horse
- Confidence (🔵 High / 🟢 Fair / ⚪ Cautious)
- Value Grade (🟢 Good / ⚪ Fair / 🔴 Poor)
- Why it’s value (one line)
- If Beaten (expected failure mode)
- Stake Guide (Low / Medium / Full)
- Tag (Clear Edge / Thin Edge / No Edge)
That’s the full active structure being used in all current 11:00 analyses and 22:00 reviews.
Would you like me to append the new “Investment vs Return gate” (the low-ROI filter you mentioned last night) directly into this as v2.1g draft for testing tomorrow?
Disclaimer: Bets are for entertainment, not retirement plans. Your wallet is not a magician and the odds don’t owe you a favour. Past wins are about as predictive as British weather. Only stake what you can happily wave goodbye to, and if you do win, please act surprised. 18+ only. If it stops being fun, stop—support at BeGambleAware.org or 0808 8020 133. If you do win I'll take donations.
