Chat GPT & Generative AI tools

A place to discuss anything.
Post Reply
User avatar
Kai
Posts: 5980
Joined: Tue Jan 20, 2015 12:21 pm

Kai wrote:
Fri Feb 09, 2024 3:02 pm
Bard rebrands as Gemini : https://blog.google/products/gemini/bar ... anced-app/
Not much has changed, it consistently gets basic factual information completely wrong

As a search engine it's pretty much useless at this point
User avatar
Archangel
Posts: 1969
Joined: Thu Jun 27, 2013 3:03 pm
Location: Polo Lounge, Beverly Hills Hotel

I see Gemini has a paid subscriber option called Gemini Advanced. I wonder does that get more answers right :lol:
Fugazi
Posts: 128
Joined: Wed Jan 10, 2024 7:20 pm

Personally I think GPT is brilliant to combine with bet angel.

The main thing is to remember it has a 30 message memory. So make sure you're very clear what you want it to do in the first message.

Break tasks down into smaller questions you want answers to.
User avatar
napshnap
Posts: 1187
Joined: Thu Jan 12, 2017 6:21 am

I have a misgiving that if a llm is just a tool for averaging (choosing the most probable answer) of all human-generated data, then using these llms and continuous injection of llms-generated data into human-generated data will average the data/will cutoff outliers, degenerate the data, possibilities to evolve.
User avatar
firlandsfarm
Posts: 2675
Joined: Sat May 03, 2014 8:20 am

Fugazi wrote:
Sat Feb 10, 2024 11:17 pm
Personally I think GPT is brilliant to combine with bet angel.

The main thing is to remember it has a 30 message memory. So make sure you're very clear what you want it to do in the first message.

Break tasks down into smaller questions you want answers to.
Hi Fugazi, do you mean a direct digital link or asking questions of GPT and using the responses to set up BA?
Fugazi
Posts: 128
Joined: Wed Jan 10, 2024 7:20 pm

firlandsfarm wrote:
Thu Feb 15, 2024 2:31 am
Fugazi wrote:
Sat Feb 10, 2024 11:17 pm
Personally I think GPT is brilliant to combine with bet angel.

The main thing is to remember it has a 30 message memory. So make sure you're very clear what you want it to do in the first message.

Break tasks down into smaller questions you want answers to.
Hi Fugazi, do you mean a direct digital link or asking questions of GPT and using the responses to set up BA?
I meant asking it questions

For example i gave it some data and started with "find a winning strategy"

Its first strategy was to pick all the horses that won...

So i then had to explain why thats an idiotic suggestion and explained more specifically the kind of thing im trying to do.

In the end it found something quite clever. A lot of back and forth.
User avatar
wearthefoxhat
Posts: 3173
Joined: Sun Feb 18, 2018 9:55 am

Fugazi wrote:
Thu Feb 15, 2024 7:14 am
firlandsfarm wrote:
Thu Feb 15, 2024 2:31 am
Fugazi wrote:
Sat Feb 10, 2024 11:17 pm
Personally I think GPT is brilliant to combine with bet angel.

The main thing is to remember it has a 30 message memory. So make sure you're very clear what you want it to do in the first message.

Break tasks down into smaller questions you want answers to.
Hi Fugazi, do you mean a direct digital link or asking questions of GPT and using the responses to set up BA?
I meant asking it questions

For example i gave it some data and started with "find a winning strategy"

Its first strategy was to pick all the horses that won...

So i then had to explain why thats an idiotic suggestion and explained more specifically the kind of thing im trying to do.

In the end it found something quite clever. A lot of back and forth.
I agree there's a lot of back and forth, but it's worthwhile.

It's revolutionised my excel sheets and formulas. I can accurately rate a horse-race with its' own tissue price, create a tick box post-data presentation, and predict a race after its' swept through all the key data in micro seconds. Makes Bill Benters 'data model redundant...(yeah right...not quite)

The main reason for producing it, is to see where the early value is for backing/laying.

The bookies seemed to have gotten lazy by following each other like sheep and relying on banning the winning punter to avoid losing. If the punter does win, then they find a reason not to pay!
Fugazi
Posts: 128
Joined: Wed Jan 10, 2024 7:20 pm

wearthefoxhat wrote:
Thu Feb 15, 2024 8:47 am
Fugazi wrote:
Thu Feb 15, 2024 7:14 am
firlandsfarm wrote:
Thu Feb 15, 2024 2:31 am

Hi Fugazi, do you mean a direct digital link or asking questions of GPT and using the responses to set up BA?
I meant asking it questions

For example i gave it some data and started with "find a winning strategy"

Its first strategy was to pick all the horses that won...

So i then had to explain why thats an idiotic suggestion and explained more specifically the kind of thing im trying to do.

In the end it found something quite clever. A lot of back and forth.
I agree there's a lot of back and forth, but it's worthwhile.

It's revolutionised my excel sheets and formulas. I can accurately rate a horse-race with its' own tissue price, create a tick box post-data presentation, and predict a race after its' swept through all the key data in micro seconds. Makes Bill Benters 'data model redundant...(yeah right...not quite)

The main reason for producing it, is to see where the early value is for backing/laying.

The bookies seemed to have gotten lazy by following each other like sheep and relying on banning the winning punter to avoid losing. If the punter does win, then they find a reason not to pay!
Ive been thinking of similar..im already pretty accurate predicting greyhounds as theres not too many metrics that are significant. Just compare times/distances mostly

Horses however , seems a lot more factors involved. Jockeys / weight/ trainer/ ground changes. Too much for me to factor together at the moment. Not to mention greyhounds have only a few distances really and same field size.

Feel like horses have a lot more depth to analysis as well.. some even have spotters that watch them train. Greyhounds dont have the liquidity for people to bother analysing as deeply so dont need to try as hard predicting price
User avatar
wearthefoxhat
Posts: 3173
Joined: Sun Feb 18, 2018 9:55 am

Fugazi wrote:
Thu Feb 15, 2024 1:07 pm
wearthefoxhat wrote:
Thu Feb 15, 2024 8:47 am
Fugazi wrote:
Thu Feb 15, 2024 7:14 am


I meant asking it questions

For example i gave it some data and started with "find a winning strategy"

Its first strategy was to pick all the horses that won...

So i then had to explain why thats an idiotic suggestion and explained more specifically the kind of thing im trying to do.

In the end it found something quite clever. A lot of back and forth.
I agree there's a lot of back and forth, but it's worthwhile.

It's revolutionised my excel sheets and formulas. I can accurately rate a horse-race with its' own tissue price, create a tick box post-data presentation, and predict a race after its' swept through all the key data in micro seconds. Makes Bill Benters 'data model redundant...(yeah right...not quite)

The main reason for producing it, is to see where the early value is for backing/laying.

The bookies seemed to have gotten lazy by following each other like sheep and relying on banning the winning punter to avoid losing. If the punter does win, then they find a reason not to pay!
Ive been thinking of similar..im already pretty accurate predicting greyhounds as theres not too many metrics that are significant. Just compare times/distances mostly

Horses however , seems a lot more factors involved. Jockeys / weight/ trainer/ ground changes. Too much for me to factor together at the moment. Not to mention greyhounds have only a few distances really and same field size.

Feel like horses have a lot more depth to analysis as well.. some even have spotters that watch them train. Greyhounds dont have the liquidity for people to bother analysing as deeply so dont need to try as hard predicting price
I did look at greyhounds quite a lot a while back. Reckon there's still money to be made straight backing or laying, mainly by specialising at certain tracks that favour front runners. Produce a split to 1st bend rating and all overall time rating will short-list the contenders, then it's a question of being in the right grade compared to the other runners.

Yes, the variables in horse racing are many. The ones that are known in the form book are fairly straight forward to assess, it's the unknown ones that are tougher to account for, and usually only once the result is final. It's satisfying when it comes right.

Today; Al Suil Eile Won @ 14/1. My tissue was 5.41 taking into account its' overall profile and data points. This of course represented value, but paranoia can creep in, second guessing yourself thinking the trainer/connections aren't trying. They backed the favourite off the board down to 15/8 as though defeat was out of the question, cementing further doubt in your algorithm. That's why we use bots to take away the emotion and take the value spots when they come.

Here's a SWOT analysis I did a while back that got me started and then asking Chat GPT further questions.

SWOT.png
You do not have the required permissions to view the files attached to this post.
Fugazi
Posts: 128
Joined: Wed Jan 10, 2024 7:20 pm

wearthefoxhat wrote:
Thu Feb 15, 2024 5:21 pm
Fugazi wrote:
Thu Feb 15, 2024 1:07 pm
wearthefoxhat wrote:
Thu Feb 15, 2024 8:47 am


I agree there's a lot of back and forth, but it's worthwhile.

It's revolutionised my excel sheets and formulas. I can accurately rate a horse-race with its' own tissue price, create a tick box post-data presentation, and predict a race after its' swept through all the key data in micro seconds. Makes Bill Benters 'data model redundant...(yeah right...not quite)

The main reason for producing it, is to see where the early value is for backing/laying.

The bookies seemed to have gotten lazy by following each other like sheep and relying on banning the winning punter to avoid losing. If the punter does win, then they find a reason not to pay!
Ive been thinking of similar..im already pretty accurate predicting greyhounds as theres not too many metrics that are significant. Just compare times/distances mostly

Horses however , seems a lot more factors involved. Jockeys / weight/ trainer/ ground changes. Too much for me to factor together at the moment. Not to mention greyhounds have only a few distances really and same field size.

Feel like horses have a lot more depth to analysis as well.. some even have spotters that watch them train. Greyhounds dont have the liquidity for people to bother analysing as deeply so dont need to try as hard predicting price
I did look at greyhounds quite a lot a while back. Reckon there's still money to be made straight backing or laying, mainly by specialising at certain tracks that favour front runners. Produce a split to 1st bend rating and all overall time rating will short-list the contenders, then it's a question of being in the right grade compared to the other runners.

Yes, the variables in horse racing are many. The ones that are known in the form book are fairly straight forward to assess, it's the unknown ones that are tougher to account for, and usually only once the result is final. It's satisfying when it comes right.

Today; Al Suil Eile Won @ 14/1. My tissue was 5.41 taking into account its' overall profile and data points. This of course represented value, but paranoia can creep in, second guessing yourself thinking the trainer/connections aren't trying. They backed the favourite off the board down to 15/8 as though defeat was out of the question, cementing further doubt in your algorithm. That's why we use bots to take away the emotion and take the value spots when they come.

Here's a SWOT analysis I did a while back that got me started and then asking Chat GPT further questions.


SWOT.png
What are you basing "excellent training regime" on? Knowledge of who the top trainers are?

Also, is a slow starter not potentially a benefit if there are horses known to go out too fast?
User avatar
wearthefoxhat
Posts: 3173
Joined: Sun Feb 18, 2018 9:55 am

Fugazi wrote:
Thu Feb 15, 2024 5:35 pm
wearthefoxhat wrote:
Thu Feb 15, 2024 5:21 pm
Fugazi wrote:
Thu Feb 15, 2024 1:07 pm


Ive been thinking of similar..im already pretty accurate predicting greyhounds as theres not too many metrics that are significant. Just compare times/distances mostly

Horses however , seems a lot more factors involved. Jockeys / weight/ trainer/ ground changes. Too much for me to factor together at the moment. Not to mention greyhounds have only a few distances really and same field size.

Feel like horses have a lot more depth to analysis as well.. some even have spotters that watch them train. Greyhounds dont have the liquidity for people to bother analysing as deeply so dont need to try as hard predicting price
I did look at greyhounds quite a lot a while back. Reckon there's still money to be made straight backing or laying, mainly by specialising at certain tracks that favour front runners. Produce a split to 1st bend rating and all overall time rating will short-list the contenders, then it's a question of being in the right grade compared to the other runners.

Yes, the variables in horse racing are many. The ones that are known in the form book are fairly straight forward to assess, it's the unknown ones that are tougher to account for, and usually only once the result is final. It's satisfying when it comes right.

Today; Al Suil Eile Won @ 14/1. My tissue was 5.41 taking into account its' overall profile and data points. This of course represented value, but paranoia can creep in, second guessing yourself thinking the trainer/connections aren't trying. They backed the favourite off the board down to 15/8 as though defeat was out of the question, cementing further doubt in your algorithm. That's why we use bots to take away the emotion and take the value spots when they come.

Here's a SWOT analysis I did a while back that got me started and then asking Chat GPT further questions.


SWOT.png
What are you basing "excellent training regime" on? Knowledge of who the top trainers are?

Also, is a slow starter not potentially a benefit if there are horses known to go out too fast?


For the training regime, I focussed on Trainers record at the track and trainer form in general. Free info on the main websites.

With slow starters, that's a grey area. The pace of the race is one of those known and unknown elements that can impact the result of a race. Say you have a race with plenty of pace up front, this can set up a race for slow starter/hold up horse. Conversely, if there is no pace up front and the field dawdles along, then the slow starter will be at a disadvantage as there will likely be a mad dash at the end.

No one, apart from the trainer/jockey will know the actual tactics used, until the race is in-play. So the best guess is made on the likely pace. Many tend to overlook this one and getting the data points right will matter in the long run. Also, In-play trading is another area that, if done right, can be very profitable. The Bet Angel In-play trader is a must to help visualise the race in progress.
User avatar
Euler
Posts: 24534
Joined: Wed Nov 10, 2010 1:39 pm
Location: Bet Angel HQ

Just watched a presentation on Sora, it's amazing: -

https://openai.com/sora
User avatar
Kai
Posts: 5980
Joined: Tue Jan 20, 2015 12:21 pm

Ye it's brilliant

One giant leap for mankind, one small update for AI :|

So we can't even trust videos anymore ffs!
User avatar
Trader724
Posts: 556
Joined: Fri Dec 27, 2019 11:22 pm

Yes there are unprecedented opportunities ahead
User avatar
Euler
Posts: 24534
Joined: Wed Nov 10, 2010 1:39 pm
Location: Bet Angel HQ

Nvidia booming!

Nvidia’s Data Center business is booming, up more than 400% since last year to $18.4 billion in fourth-quarter sales

https://www.cnbc.com/2024/02/21/nvidias ... sales.html
Post Reply

Return to “General discussion”