The art of buying a yearling that goes on to be a Stallion

[Chief Stipe]

How does AI differentiate between fact and opinion?

[hesi]

Would the benefit of AI to horse racing and wagering, being able to use its processing power to for example analyse the many factors of all horses in a race, such as weight, barrier, sectionals in previous races, jockey, track conditions, track, trial form, unlucky runs etc etc etc to determine the most likely winner.

Much the same as many punters do manually.

When a $30 outsider wins, you can quite often review its form and find a reason why it was going to run so well

[Gammalite]

12 minutes ago, Chief Stipe said:

How does AI differentiate between fact and opinion?

The algorithms are designed by human at the moment . 

Fact and Opinion are common are common in daily life. especially on forums lol 😂 even at the start of a Person's day you could say there are facts and opinions in play from the get -go. Say even about the weather (It's not raining currently ) so that's a fact ,  and opinion might be that it might rain later on. another example is > the traffic is heavy now as people head off to work is a Fact . a slow trip in to the city. but later on it might clear and I'll get into town a lot quicker , is just an opinion though , until it actually happens. totally BASED off previous trends.

So A.I will use Past events as FACT , as they have happened already. 

and coming events as OPINION , with a degree of inaccuracy. The event hasn't happened yet. So a big rate of inaccuracy will always be there, based on the conditions surrounding the event being described.

such as selecting a yearling to be a stallion like in this thread ?? . a near impossible Task. we all know this based on previous trends. and like Most things in the amazing Racing field , is just a matter of Opinion . 

Punting is a great topic for A.I to handle though Chief. You see so many people on forums that have the 'Ideal' strategy in their minds and even shout down others that approach it in a different method. It's quite funny how volatile people get about it actually 😂  Forgetting there are lots of methods. And results are extremely VARIABLE.

I copped abuse on Hesi's forum one time, for suggesting weights and barrier draws are part of the variables in your chance of winning a race. Think those blokes need some A.I to help them along a bit 😎🤣 They have a lot of trouble differentiating between Fact and Opinion , just like the A.I . 

All you can do is go off past trends = (facts) to form future Analysis = (opinion) just like reading the form guide Chief. 

 

 

[Chief Stipe]

9 minutes ago, hesi said:

Would the benefit of AI to horse racing and wagering, being able to use its processing power to for example analyse the many factors of all horses in a race, such as weight, barrier, sectionals in previous races, jockey, track conditions, track, trial form, unlucky runs etc etc etc to determine the most likely winner.

Only if that data was consistently and accurately collected.

10 minutes ago, hesi said:

Much the same as many punters do manually.

Many punters do it programmatically.  The advantage to them of AI is machine learning and the ability of higher performance computers to tweak the punters algorithm.

12 minutes ago, hesi said:

When a $30 outsider wins, you can quite often review its form and find a reason why it was going to run so well

I'm disappointed @hesi that someone of your science background would fall for observational bias.  The fact your algorithm didn't pick it up in the first place would suggest that you have subjectively chosen a factor to justify the result.  AI would probably work best when ALL data was available but in racing the data/information available is imperfect.

[curious]

Here's an AI view on that question.

AI differentiates between fact and opinion using a combination of techniques, but it’s not perfect. Here’s how it generally works:

---

1. Source Analysis

• Fact: Comes from verifiable, authoritative sources (e.g., official statistics, peer-reviewed research, reputable news).
• Opinion: Often found in editorial content, blogs, or statements attributed to individuals or groups.
• AI checks metadata, domain reputation, and context clues to classify the source.

---

2. Linguistic Cues

• Facts often use objective language: numbers, dates, measurable claims (e.g., “The NZ Derby was won by Zonda in 1997”).
• Opinions use subjective markers: “I think,” “should,” “best,” “amazing,” or emotionally charged adjectives.
• AI models are trained to detect these patterns.

---

3. Cross-Verification

• AI compares statements against trusted knowledge bases or multiple sources.
• If a claim is consistent across reliable sources, it’s likely factual.
• Divergent or speculative statements are flagged as opinion.

---

4. Confidence Scoring

• AI assigns a probability score based on evidence strength.
• High confidence = fact; low confidence or conflicting sources = opinion or uncertainty.

---

5. Context Awareness

• Some statements mix fact and opinion (e.g., “New Zealand played well because they dominated possession”).
• AI splits the factual part (“dominated possession”) from the subjective part (“played well”).

---

Limitations

• AI can misclassify nuanced language or sarcasm.
• It relies on training data and source integrity—bias in sources can affect classification.

[curious]

5 minutes ago, Chief Stipe said:

I'm disappointed @hesi that someone of your science background would fall for observational bias.  The fact your algorithm didn't pick it up in the first place would suggest that you have subjectively chosen a factor to justify the result.  AI would probably work best when ALL data was available but in racing the data/information available is imperfect.

I agree with that. Retrospectively giving undue weight to a factor that doesn't deserve it is a recipe for disaster. That's where a lot of punter and even trainer believed myths are founded and perpetuated but fraught.

Edited 1 hour ago by curious

[Chief Stipe]

5 minutes ago, Gammalite said:

The algorithms are designed by human at the moment . 

Yes and no - the intial algorithms to start AI - machine learning are designed by humans.  However the computational power is being used by machines to rewrite those algorithms.  

 

7 minutes ago, Gammalite said:

So A.I will use Past events as FACT , as they have happened already. 

and coming events as OPINION , with a degree of inaccuracy.

No - if AI says there is a 90% chance that it will rain within an hour that isn't opinion it is a prediction based on past observations and recorded data i.e. facts.  

The worst thing about AI at the moment is that it is much of what is produced doesn't come with its own measure of accuracy and variance.  So 

[Chief Stipe]

8 minutes ago, curious said:

Limitations

• AI can misclassify nuanced language or sarcasm.
• It relies on training data and source integrity—bias in sources can affect classification.

Damn so @Huey 's posts could still be counted!!

[hesi]

1 minute ago, curious said:

I agree with that. Retrospectively giving undue weight to a factor that doesn't deserve it is a recipe for disaster. That's where a lot of punter and even trainer believed myths are fraught.

Factors you may have overlooked or not given sufficient weighting to.

Tweaking your evolving algorithm or system or whatever it is you use.

I don't use one because I rarely bet these days, but still an interesting topic

[Gammalite]

8 minutes ago, Chief Stipe said:

No - if AI says there is a 90% chance that it will rain within an hour that isn't opinion it is a prediction based on past observations and recorded data i.e. facts. 

I don't know about that? . I would say that any PREDICTION for anything , is just an OPINION.

until it happened as that it is now in 'past tense'. so therefore becomes a FACT.

The Weather Office should have data to accurately predict the weather most days but wow. the inaccuracies are still there in bulk. People complain every day.

As they do in horse racing every day too. Not getting the results they predicted.  Racing is full of inaccurate Predictions of finishing order, how a race will be run, any number of things really. All just opinion. 

your Rain example shows the Inaccuracy of A.I in all it's glory. It only has to rain in say 8 of a total of 10 hours of predictions (based on the recorded past data you mentioned) and it's PREDICTION is Immeadiately 10% incorrect. 

therefore it was just an OPINION of what might happen .and just like any horse race. an inaccurate opinion on any unknown number of occasions.  

[Chief Stipe]

8 minutes ago, Gammalite said:

I don't know about that? . I would say that any PREDICTION for anything , is just an OPINION.

An opinion is subjective.  A quantifiable prediction based on data past and present with an estimate of probability is different.  If I say that based on the data I have I predict in will rain within one hour and that prediction has a 90% chance of happening.  Then I can measure the outcome against that prediction.  I'm saying that 9 times out of 10 I'll be right.

But if I say Gammalite will win the first at Addington but offer no data supporting my selection and probability of it happening then it's just a hunch i.e. a subjective opinion.

[Chief Stipe]

14 minutes ago, Gammalite said:

The Weather Office should have data to accurately predict the weather most days but wow. the inaccuracies are still there in bulk. People complain every day.

However the forecast models are significantly more accurate than they were 40 years ago.  Their accuracy increases the closer you get to the day or hour.

My partner uses the YR weather forecasting model and often I get told that YR says it will be such and such on a specific race day in two months time.  I invariably laugh and no matter how hard I try to explain that we have a maritime and very variable weather pattern I cannot move her.  As I say to her two weeks out most weather forecasts reach 70% probability or higher.  A week out and they are getting close to 90%.   If you are deciding what to wear I suggest she waits until the last 48 hours.  If she wants to choose earlier I suggest that she at least waits until two weeks out and if there are bargains before then buy a hot and a cold one.

[Chief Stipe]

22 minutes ago, Gammalite said:

your Rain example shows the Inaccuracy of A.I in all it's glory. It only has to rain in say 8 of a total of 10 hours of predictions (based on the recorded past data you mentioned) and it's PREDICTION is Immeadiately 10% incorrect. 

No that isn't how probability works.  

[Gammalite]

7 minutes ago, Chief Stipe said:

An opinion is subjective.  A quantifiable prediction based on data past and present with an estimate of probability is different.  If I say that based on the data I have I predict in will rain within one hour and that prediction has a 90% chance of happening.  Then I can measure the outcome against that prediction.  I'm saying that 9 times out of 10 I'll be right.

But if I say Gammalite will win the first at Addington but offer no data supporting my selection and probability of it happening then it's just a hunch i.e. a subjective opinion.

Fair enough. I would still contest the 'Laws of Probability' in mathematics and even calculating for an event to occur,  are slightly different to the PREDICTION and OPINION of an outcome.  but they do all inter-twine I guess. 

Gammalite didn't win at Addington in 5 attempts as the Brodster had the Canterbury locals 'fired up' far too much 😎 so the 'Prediction' of a NZ Cup win (or any Addington Win didn't happen , and the 'Opinion' of that is well  > "That Sucks'. 🤣 

but the 'Laws of Probability' of winning the NZ Cup off the Handicap (on 2 occasions)  given to him suggested he would Not win the race anyway. 

 

[curious]

A prediction to me, is an informed opinion, NOT a fact, as it hasn't happened.