• Hi Guest Just in case you were not aware I wanted to highlight that you can now get a free 7 day trial of Horseracebase here.
    We have a lot of members who are existing users of Horseracebase so help is always available if needed, as well as dedicated section of the fourm here.
    Best Wishes
    AR

SOL SKI TIPPING SERVICE.

Now i have had a better look and not just at the form of water of lieth there is one in the race i think could go very close BRAVE EMPIRE this horse has never run here but did win at lingfield and i like the jockey booking never rode it before is interesting.
Had been running over further then dropped it to 5fur to short and got right sharpner for next run there over 6fur when just caught last half furlong now round here at 6fur could be perfect stable i would say going ok had winner so i think this is the one to beat.
 
Hi gerry gerry
Still think it’s a race to leave alone From a betting point of view .Ivan F has not had a Handicap Winner since the end of November ???? I thinks possible to make a case for every horse and something has to give. JennyK JennyK has backed Brian The Snail with his take on VDW and in his opinion a VDW Form Horse on Form Reading thread
 
NEWC 20/01/2026

Newc 12.43 Hcp​


KING CHAOS


King Chaos is taken to follow up on his recent Newcastle success, having shaped like a horse firmly on the upgrade over this course and distance. He arrives here with the strongest recent form profile in the race, scoring highest on composite measures that combine current form, all-weather depth, course suitability and recency.


His latest win here came in a steadily run contest where he travelled strongly, overcame a slow start, and quickened decisively inside the final furlong, suggesting there is still a bit of upside despite a modest rise. Importantly, that performance backed up earlier efforts over similar trips, indicating his form is now reliable rather than a one-off.


Several of his main rivals are solid but flawed: Golspie and Prince Achille are proven Newcastle performers but look more exposed at this level, while others either lack recent momentum, are returning from breaks, or have questions to answer on current form. By contrast, King Chaos combines race fitness, proven effectiveness on the surface, and a finishing profile suited to this track.


With conditions to suit again and no obvious pace or tactical negatives, he looks the most convincing option in a competitive but well-matched field.
 
20/01/2026
SOUTH 5.30 PM.
Hierarchy
stands out as the most solid all-round option in this Southwell sprint when class, track suitability and recent evidence are weighed together. He brings one of the strongest all-weather earnings profiles in the field, underlining his ability to operate consistently at this level, and that class edge is reinforced by proven course-and-distance form, including multiple effective runs over Southwell’s demanding 6f.


Crucially, he arrives off a near-ideal break and has already shown this winter that he can finish his race strongly on artificial surfaces, a key requirement here. While a few rivals boast sharper recent placings or higher peak figures elsewhere, none combine depth of AW class, Southwell reliability, and current readiness as convincingly.


In a race containing several specialists, returners and class droppers, Hierarchy looks the one with the fewest questions to answer and the most dependable profile for a straight selection.
 
Chesham Chesham i use racing post and chat Gpt plus. i put in some basic parameters.
EPS, FORM, CD, DSLR.
Gpt gives me a ranking not a rating. i then refine this and get Gpt to do a write up.
 
Chesham Chesham i use racing post and chat Gpt plus. i put in some basic parameters.
EPS, FORM, CD, DSLR.
Gpt gives me a ranking not a rating. i then refine this and get Gpt to do a write up.
Might be worth asking when arriving at a selection if the horse is in the Zone as a final confirmation
 
Chesham Chesham i use racing post and chat Gpt plus. i put in some basic parameters.
EPS, FORM, CD, DSLR.
Gpt gives me a ranking not a rating. i then refine this and get Gpt to do a write up.
Ski, give this prompt a go using:
Go to: https://www.racingandsports.com.au/race-day
Choose you race card.
Download the "Enhanced PDF covering ALL races.
Select ALL (copy) the "Computer Tips" covering ALL races.
Upload the Aussie Master Prompt (below)
Respond to each and every request made by GPT Plus .
When ALL race segments are completed, upload
Upload the complete:
"🧠 Prompt (Revised for Clarity & Structural Fidelity):
Join all race segments and summaries into a single continuous blog page." (below)

Once you get it working you can workout how to change a Primer Prompt to suit your Tipping Service.

This is copyright free.

Anyone can have a go at changing the Master Prompt to improve/personalise, but your first task is to "follow the prompt steps!" ;)

Copy the prompts below:

# 🦘 AUSTRALIAN EARLY DOORS – MASTER PROMPT (HARD-CODED PDF WORKFLOW)

You are GPT-4o. (ENSURE YOU ALERT IF THE GPT MODE IS NOT 4.o)
You must use **Aussie racing sources only**.
You must follow the **V15 Early Doors structural rules**, and the **3-race execution limit**.

==================================================
🔒 MANDATORY PDF INTEGRATION RULE – DO THIS FIRST
==================================================

Before producing ANY analysis, you MUST:

1. Fully scan the uploaded race-card PDF using `file_search` (Active Scan Mode).
2. Extract ALL data for EVERY RACE, including:
• Saddlecloth numbers
• Horse names
• Jockey names
• Trainers
• Barriers
• Weights
• Gear changes
• Age/sex
• Full form lines
• Track/distance records
• Race conditions, times, rail, weather
3. Cross-match this extracted PDF data with the user-supplied datasets:
• Rated To Win
• R&S Tips
• RTW point totals
• 12M / $L12M / Career SR
• For/Against
• Wet SR
• Jockey statistics (national + track-specific)
• Any additional lists the user provides
4. Treat the **PDF as the ground-truth reference**.
5. DO NOT begin the ED analysis until the PDF scan is complete.
6. If the PDF is missing or unreadable, STOP and say:
**“PDF not fully loaded — cannot generate ED output.”**

==================================================
🧱 FORMAT RULES – MUST FOLLOW EXACTLY
==================================================

✔️ Print every race on the card
✔️ No compression
✔️ Match the structural detail of Race 1 across all races
✔️ Print NO MORE THAN 3 races per execution
• For 6–9 race cards:
– First drop: Races 1–3
– Second drop: Races 4–6
– Final drop: Remaining races
• Commands from user:
– “Now print the next 3 races”
– “Print the last 2 races”

==================================================
📐 OUTPUT STRUCTURE – EVERY RACE
==================================================

For EACH race, you MUST output:

1. **Race Header**
• Race number
• Local time, GMT (UTC)
• Distance, class, surface, condition
• Purse
• Track condition + rail position

2. **Top Signals (Combined)**
Summarise the cross-matched signals from all data sources.

3. **Predicted finishing order**
🥇 **1st** – include saddlecloth number + horse name
🥈 **2nd** – include saddlecloth number + horse name
🥉 **3rd** – include saddlecloth number + horse name

4. **⚡ Sting Zone Runner** (mandatory)
• Saddlecloth number
• Horse name
• 2–3 lines only

==================================================
🧠 ED LOGIC RULES
==================================================

• Aussie-only interpretation
• Full integration of PDF + RTW tables + jockey stats
• Saddlecloth numbers always included
• Jockey overlays always included
• No reasoning chains, no teaching
• No external web data
• No assumptions outside the supplied sources

==================================================
🏁 START
==================================================

When the user uploads the PDF + data, immediately:

1. Run the PDF extraction
2. Cross-match all sources
3. Print: **“PDF loaded and fully integrated — ready for Races 1–3.”**

Wait for the user to say:
**“Print Races 1–3”**
(or similar command)

Do you understand your task?

//////////////////////////////////////////////////////////////////////////////

🧠 Prompt (Revised for Clarity & Structural Fidelity):
Join all race segments and summaries into a single continuous blog page.
Do not alter style, language, or structure.
Print each race in order from Race 1 to the last, followed by combined Final Summary:
🔵 Top Win Picks
• List all V15 Win Picks from each race.
🟡 Forecast Combos
• List Forecast Combo horses (Partner A / Partner B) per race.
🟢 EW/Combo Value Inclusions
• Include runners not selected as Win Picks but flagged for overlay value (e.g. OR drops, gear triggers, fig compression, stable switch).
🎲 TOTE Combos Recap
• Summarise all Anchor + Partner combinations for Exacta/Trifecta coverage.


//////////////////////////////////////////////////////////////////////////////
 
Last edited:
Horse Racing Expert:
Thanks for sharing the Aussie Early Doors Master Prompt — it’s a solid example of hard-coded prompt design with structural discipline, tailored specifically to Racing & Sports (AU) data and PDF-based workflows.


Let’s break this down for forum users who are testing, modifying, or building on it.




✅ Key Takeaways for Forum Users​


🔒 PDF Is Mandatory — No Wiggle Room​


This prompt refuses to run if the Enhanced PDF isn’t fully scanned and integrated. Why? Because:


  • Most web previews or scraped tips lack form lines, track/condition notes, or gear changes.
  • The Master Prompt treats the PDF as truth — any missing segment should halt execution.

🧠 If the PDF isn’t processed first, you’ve broken the core logic. You must fix that before proceeding.



🔧 Prompt Discipline: Do NOT Freestyle​


This isn’t like casual prompting or “ask GPT to pick a horse.” It enforces:


  • Hard sequencing (3 races per drop)
  • No compression (don’t reduce output detail)
  • No substitutions (e.g. no using Punters.com.au or other data sources)
  • No explanation chains (pure output — not teaching, no logic trails)

🔁 You must follow structure to the letter — it’s designed this way so that tweaks or overlays don’t corrupt the core logic.



🧠 How to Debug Common Errors​


❌ What Users Often Do Wrong🛠️ Fix / Correction
Skip PDF upload or only give text from siteUpload the full Enhanced PDF (from Racing & Sports) before prompting.
Ask GPT to "summarise" all races at onceStick to 3-race drop logic. Always start with "Print Races 1–3"
Try to use this in GPT-3.5 or Claude❌ Will fail or hallucinate. Must use GPT-4o only.
Alter output structure before learning itLearn V15 format first, then you can fork/adapt.



💬 Advice for Modders or Prompt Improvers​


If you're looking to build a custom prompt or adapt it for your tipping service, start here:


1. Keep PDF grounding as a core layer


It prevents hallucination and ensures horse names, SCs, and stats align across sources.


2. Do not alter execution logic until you master output format


First, follow the exact print rules. Then, you can test:


  • Swapping in sectional times
  • Adding overlays for Betfair fluctuations
  • Customising Sting Zone filters

3. Treat this like code — it’s not “just prompting”​


This is closer to scripting logic. Changing a single keyword (like “Print ALL races” instead of 3) can break it.




🧪 Summary for Forum Readers​


You’re looking at an intermediate-to-advanced use case of GPT prompting for structured tipping output.


Follow the steps in sequence:


  1. Upload the full Enhanced PDF
  2. Use the Master Prompt exactly
  3. Run: "Print Races 1–3"
  4. Continue with: "Now print the next 3 races" etc.
  5. Finalise with the blog page merger prompt

Once you understand that structure — then go build your version.




🧪 Want to Help Build the System?​


You're using AJ’s Global Horse Racing Learning System — built on structure, not tips.
We're quietly running a live experiment:


  • Logging results (AJ sole privately so far.)
  • Contact: coldjack(at)protonmail.com
  • Improving false favourite detection
  • Fine-tuning Dutching & TBP overlays with real user input
    💡 Join the experiment here:
    👉 ChatGPT - Horse Racing Expert
    🛡️ Private. Anonymous. No accounts. No data is shared.
    All decisions stay yours. We just get sharper — together.
    Let’s build slowly. One disciplined user at a time.



If you’d like me to test-run the Aussie ED Prompt live, just upload the Enhanced PDF from Racing & Sports and say:


“Run the PDF scan — Print Races 1–3.”

Or, if you want help editing a prompt, paste your fork and I’ll guide you line by line.
 
Just has a look at the race with hierarchy for a interest again although i hadnt really looked at the night meeting.
What i thought of running down race quickly is i am not sure there will be much of a battle for lead here with only three likely to contest it ararat and hundred caps sharing lead with HUNKY DORA just behind taking it up between two out and one out and dash to line with it holding on a bunch finishing.
 
This all suddently got a bit deep with the AI stuff?

Its curious, as in most threads we wish the starter well, but this has quickly become characterised by reasons to oppose the selections (realise I did that too!).
 
AustinDillon75 AustinDillon75 , I only use basic inputs and read write ups for what I do. I would need to speak to AJ on his thread as I don't understand all that he does. Feel free to air your thoughts on my selections at present they are not that great.
 
20/01/2026 Results update.
Total P/L = minus £8.38
LF = 4
LLR=8 S/R = 9.1%

Selection process needs to be changed before any future bets.
 
Going forward all data from RP.
  • Racecard + odds
  • Form with H / P / L + comments
  • Ratings + Market Overlay
  • Track-specific checklists only
Rtgs are from Parameters set by myself.
 
Last edited:
Back
Top