
- Value comes from comparing your probability with the odds: the AI does the calculation, not the estimate.
- The most useful prompt is the one that challenges your number: ask it to play devil’s advocate.
- If your value disappears with 3 points of error, it is fragile value and it is better to pass.
Save this article: the next time a set of odds looks high to you, come back, copy the prompts in order and pick up where you left off.
Your progress is saved in this browser.
Which prompts work for spotting value bets?
An AI does not find value on its own.
Value appears when your probability is higher than the one the odds pay.
The AI does not have that probability: you build it.
What it does well is the boring part.
It turns odds into probabilities, calculates expected value and looks for gaps in your reasoning.
That is why this article splits the work into five prompts, one per task.
All the prompts use parameters in [BRACKETS].
Replace each one with your own data before you paste.
If you leave a bracket unfilled, the AI fills it with whatever comes to mind.
Golden rule: if you ask “is there value in this match?”, the AI makes up a probability so it can answer.
Never accept a probability you did not build yourself.
Ask it for calculations and objections, not a magic number.
If the concept is new to you, start with what value betting is and value bets explained step by step.
We do not repeat it here: we go straight to the prompts.
For prompts for a specific market there is the index of ChatGPT prompts for every betting market.

Before you paste: three rules so the AI answers better
- Paste the odds yourself: the AI does not see your bookmaker and does not know what time you looked at them.
- Write your probability before asking: if the AI gives its opinion first, your number sticks to its own.
- Ask for the calculations step by step: one operation per line can be checked in a minute.
Web search and reasoning mode depend on each AI.
Features change: check in the app what your plan includes.
No AI bets for you or knows the result, and all of them can make up data.
Prompt 1: turn the odds into implied probability and remove the margin
This is the starter prompt.
It tells you what probability the market is paying, which is the number you will compare yours against.
If you do not remove the margin, that probability comes out inflated.
I have these decimal odds for [HOME TEAM] vs [AWAY TEAM], seen at [BOOKMAKER] at [TIME]: home [ODDS 1], draw [ODDS X], away [ODDS 2]. Calculate the implied probability of each result as 1 divided by the odds. Add the three and tell me how far over 100% they go: that is the bookmaker’s margin. Then divide each probability by that sum to get the probability without the margin. Show each operation on one line, with two decimals. Do not give opinions about the match and, if I did not give you a piece of data, write “I don’t know”.
What it returns
Three probabilities with margin, the margin and three probabilities without margin that add up to 100%.
Example: odds of 2.10, 3.40 and 3.60 give 47.6%, 29.4% and 27.8%, which add up to 104.8%.
Without the 4.8% margin they become 45.4%, 28.1% and 26.5%.
How to verify it
Redo one of the divisions with the odds converter.
Check that the three probabilities without margin add up to 100%.
If you want to understand the calculation, there are the articles on implied probability and on what the bookmaker’s margin is.
Prompt 2: build your estimate and have the AI challenge it
This is the prompt that decides everything.
You paste the data and your probability, and the AI looks for reasons to knock it down.
You do not ask it to estimate for you: you ask it to take the other side.
This is the data I gathered for [HOME TEAM] vs [AWAY TEAM] on [DATE] in [COMPETITION]: [PASTE YOUR DATA WITH ITS SOURCE]. From that I estimate that [RESULT] has a probability of [MY PROBABILITY]%. Play devil’s advocate. Tell me: 1) which of my data is weak or old, 2) which important piece of data is missing, 3) what argument someone who thinks the opposite would use and 4) whether my number is far from the market’s probability without margin, which is [PROBABILITY WITHOUT MARGIN]%. Do not give me your own probability or a betting verdict. If you use a piece of data I did not give you, cite the source, and if you cannot verify it write “I don’t know”.
What it returns
A list of objections to your estimate.
The most useful is usually the fourth: how far you are from the market.
If your number is ten points above the market, you are most likely missing information.
How to verify it
The objections are opinion: take them as questions, not as answers.
If the AI brings a new piece of data, open the source before you move your number.
Write down your probability before and after the discussion, with the reason for the change.
The market is data too.
Heavily bet odds sum up the opinion of many people with a lot of information.
Moving away from it takes a concrete reason, not a hunch.
Prompt 3: calculate the expected value and the minimum odds
With your probability already challenged, the calculation arrives.
Expected value is your probability times the odds, minus 1.
The minimum odds are 1 divided by your probability.
The odds for [RESULT] in [HOME TEAM] vs [AWAY TEAM] are [ODDS] and my estimated probability is [MY PROBABILITY]%. Calculate, one operation per line: 1) the expected value per unit staked, as probability × odds − 1, 2) the minimum odds I need, as 1 divided by my probability, and 3) the break-even probability of the odds, as 1 divided by the odds. Then tell me how many percentage points there are between my probability and the break-even one. Express the expected value in money too for a stake of [STAKE]. Do not change my probability and do not tell me whether it is worth betting.
What it returns
Three numbers and a distance.
Example: with 50% and odds of 2.10, the expected value is +5.0% and the minimum odds are 2.00.
The break-even probability is 47.6%, so your cushion is 2.4 points.
| Your probability (example) | Odds | Expected value | Minimum odds | Reading |
|---|---|---|---|---|
| 47% | 2.10 | −1.3% | 2.13 | No value |
| 50% | 2.10 | +5.0% | 2.00 | Value, if your number is good |
| 53% | 2.10 | +11.3% | 1.89 | Value with a cushion |
| 40% | 2.40 | −4.0% | 2.50 | No value |
The numbers in the table are an example to show the calculation.
The first three rows use the same odds and only your probability changes.
Six points of difference in your estimate separate a bad bet from a very good one.
How to verify it
Repeat the calculation in the expected value calculator.
The full formula, with Kelly included, is in the value bet calculator.
If you want to audit the reasoning line by line, that flow is in how to bet with DeepSeek.
Prompt 4: what happens if your probability is 3 points off
Your probability is an estimate, not a measurement.
This prompt moves your number up and down and looks at what happens to the value.
It is the test that separates firm value from fragile value.
My estimated probability for [RESULT] is [MY PROBABILITY]% and the odds are [ODDS]. Build a table with five rows: my probability minus 3 points, minus 1.5, no change, plus 1.5 and plus 3. In each row calculate the expected value as probability × odds − 1. Tell me from which probability the expected value turns negative. Show the operations and do not add opinion about the match.
What it returns
A table of five rows and a breaking point.
In the example of 50% at odds of 2.10, with 47% the value is already negative: −1.3%.
An error of 3 points is enough to wipe out that value.
How to verify it
The breaking point has to match the break-even probability from prompt 3.
If it does not, there is a calculation error in one of the two answers.
Ask yourself honestly whether your estimate is accurate to within 3 points.

Prompt 5: the checklist before betting on value
This is the closing prompt.
It does not analyze the match: it reviews your work.
You ask it to look for gaps before you put money down.
This is my value analysis for [HOME TEAM] vs [AWAY TEAM] on [DATE]: [PASTE ODDS, PROBABILITY WITHOUT MARGIN, YOUR PROBABILITY, EXPECTED VALUE AND SENSITIVITY TABLE]. Act as a critical reviewer. Tell me: 1) whether the odds I used may have changed since [TIME], 2) which piece of data in my estimate is not verified, 3) whether the value survives an error of 3 points, 4) whether the stake of [STAKE] respects my rule of [STAKING RULE] and 5) what bias I may have with this team. Do not give me a betting verdict. Just flag risks.
What it returns
A short list of risks.
If point 3 fails, the value was fragile.
If point 1 fails, look at the odds again before doing anything.
How to verify it
Tick off each risk you resolved.
The ones you cannot resolve stay as a reason to skip the match.
A good value analysis often ends in not betting.
How to use the 5 prompts in order: step by step
Common mistakes when looking for value with an AI
- Asking the AI for the probability and calculating the value with that made-up number.
- Comparing your probability with the implied one without removing the margin.
- Using old odds: value is calculated with the odds you actually bet at.
- Adjusting your probability until the value comes out positive.
- Judging the method by ten bets: value shows over hundreds.
The last mistake is the most expensive.
A bet with value loses many times, and one without value wins many times.
That is why it is worth reading how many bets you need to know whether you are winning.
What Gambeta does with value, no promises
Gambeta publishes predictions with its own model.
Each one is recorded with its odds before the match.
You can review it afterwards in the public record, win or lose.
If you want to see how it works, go to the guide to AI for value betting, with free predictions from Gambeta’s AI every day.
A prediction is not a promise.
No value bet is safe.
For adults 18 and over only.
Bet only money you can afford to lose, set a limit before you start and stop if you stop enjoying it.
More at responsible betting.
What to read next
Value depends on the odds, and the odds move.
The next step is 5 prompts for reading odds movement.
And to know in which markets to look, there is where to find value in football betting.
When you have bets on record, review them with the prompts to audit your betting history.
If your market is the match result, continue with the prompts for 1X2.
And if it is goals, with the prompts for over and under.
Frequently asked questions
Can ChatGPT find value bets?
Not on its own.
The AI does not know the real probability of a match and does not see your bookmaker’s odds.
It can do the calculations and challenge your estimate, which is the part that decides whether there is value.
How is the value of a bet calculated?
You multiply your probability by the decimal odds and subtract 1.
Example: 50% at odds of 2.10 gives 0.50 × 2.10 − 1 = +5.0%.
If the result is negative, there is no value at those odds with your number.
Which prompt should I use to spot a value bet?
Start with the one that removes the margin from the odds.
Then use the one that challenges your estimate and the expected value one.
Finish with the sensitivity prompt and the checklist.
Where do I get my probability from?
From data you can check: results, goals, absences, home advantage and schedule.
You write it yourself before asking the AI for an opinion.
If you cannot explain where your number comes from, it is not an estimate.
How high does the expected value have to be to bet?
There is no magic number.
What matters is that the value stays positive even if your probability is a few points off.
A small value with a doubtful estimate is noise.
Does a value bet always win?
No.
A bet with positive expected value loses many times.
Value only shows over hundreds of bets, and betting always carries the risk of losing money.