Chat GPT & Generative AI tools

A dedicated space to explore and discuss how cutting-edge technologies are reshaping sports trading and betting strategies. Collaborate on innovative ideas, and stay ahead of the curve in a rapidly evolving landscape.
Post Reply
User avatar
Euler
Posts: 26245
Joined: Wed Nov 10, 2010 1:39 pm

firlandsfarm wrote:
Wed Jul 05, 2023 9:42 pm
Euler wrote:
Wed Jul 05, 2023 3:20 pm
ChatGPT is a large language model, it's not a statistical model. Lots of people jumping on the bandwagon to promote 'AI' services just shows how out of touch they are IMHO.
Agreed Euler, I came to that conclusion within 10 minutes of playing with it re statistical questioning so that takes me back to my original post ... are there many different AI bots out there or are they all ChatGPT based with a different front door for entry?
I get a load of 'new' things dropped in my inbox each morning, but they are just layered on top of the ChatGPT API, they are just a variation on a theme. Most are pretty poor.

We are in the upward curve of the hype cycle it seems.

Plenty of good apps out there, but few and far between. Some good stuff is being built and it is a game changer, but you have to see through the hype.
User avatar
ShaunWhite
Posts: 10374
Joined: Sat Sep 03, 2016 3:42 am

firlandsfarm wrote:
Thu Jul 06, 2023 9:15 am
ShaunWhite wrote:
Thu Jul 06, 2023 12:35 am
But there's a distinction between consumer AI like ChatGPT and 'pro' AI you train yourself. ChatGPT is getting headlines but the more interesting area for us is the cheap pro tools that people can feed with verified data and data not pubilcally available, like Betfair data.
Agreed, and that was my reason for the question i posed a few posts ago ...
Azure OpenAi Service or Microsoft Business Intellegence.

Quant traders have been using AI for several years. But before you go down the rabbit hole so far it's no more effective than traditional statistical analysis techniques, mainly because under the skin that's what it's doing because stats are like physics, there's no new rules and it is what it is.
foxwood
Posts: 429
Joined: Mon Jul 23, 2012 2:54 pm

Beware also the hype creeping in. Someone just showed me a professional drawing program and the person showing me was excited that the new version used AI. The package had always had a feature for doing a specific type of "artistic" manipulation of a picture. Over the years and the different versions it seems to have been called: "automatic", "optimized", "smart" and so on with each new edition. The latest version now prefixes this menu item with the phrase "AI" instead - all the variations do the same basic thing as far as I can tell !
User avatar
Euler
Posts: 26245
Joined: Wed Nov 10, 2010 1:39 pm

ShaunWhite wrote:
Thu Jul 06, 2023 12:58 pm
Quant traders have been using AI for several years. But before you go down the rabbit hole so far it's no more effective than traditional statistical analysis techniques, mainly because under the skin that's what it's doing because stats are like physics, there's no new rules and it is what it is.
I did a deep dive with a Neural net about 15 years ago. But it didn't end up influencing what I did based on the above, which is a good interpretation.

I'd say a lot of the current generative AI is sending people down the wrong path as a lot of what is generated is not accurate. But it takes an expert to spot that. We trained something our user guide and other advice and it invented a function we didn't have. But people are taking things as fact, although they cleary are not.

But it's still a step forward when used correctly and has other applications. I'm actually pretty excited by how quickly things are moving in the space.
Michael5482
Posts: 1693
Joined: Fri Jan 14, 2022 8:11 pm

I remember that horse racing AI Bet Beth last year was it.

Claimed to be AI driven , calculating everything down to wind speed but looked nothing more than a glorified arbing tool. Lasted 5 minutes.
User avatar
Euler
Posts: 26245
Joined: Wed Nov 10, 2010 1:39 pm

That's been in the back of head all the time on this thread - viewtopic.php?f=2&t=24382
User avatar
Euler
Posts: 26245
Joined: Wed Nov 10, 2010 1:39 pm

OpenAI just announced that code interpreter will be available at some point next week. That's going to be really helpful for non language analysis.
User avatar
Realrocknrolla
Posts: 1910
Joined: Fri Jun 05, 2020 7:15 pm

Euler wrote:
Fri Jul 07, 2023 10:42 am
OpenAI just announced that code interpreter will be available at some point next week. That's going to be really helpful for non language analysis.
AgentGPT and AutoGPT are pretty unique
User avatar
Euler
Posts: 26245
Joined: Wed Nov 10, 2010 1:39 pm

Euler wrote:
Fri Jul 07, 2023 10:42 am
OpenAI just announced that code interpreter will be available at some point next week. That's going to be really helpful for non language analysis.
They put me on it last night so I've been playing.

It's pretty neat and basically a recursive intelligent language based form of coding. Will have a good play next week.
User avatar
firlandsfarm
Posts: 3314
Joined: Sat May 03, 2014 8:20 am

Euler wrote:
Sat Jul 08, 2023 2:13 pm
Euler wrote:
Fri Jul 07, 2023 10:42 am
OpenAI just announced that code interpreter will be available at some point next week. That's going to be really helpful for non language analysis.
They put me on it last night so I've been playing.

It's pretty neat and basically a recursive intelligent language based form of coding. Will have a good play next week.
Whoosh as that flew over my head! :lol: what is a "recursive intelligent language"? :)

I just Googled it and now I'm even further from understanding it! :oops:
User avatar
firlandsfarm
Posts: 3314
Joined: Sat May 03, 2014 8:20 am

I just had the following convo with the Bing/ChatGPT ...

Me: "what was the price og gbpusd on last friday" (I didn't notice the typo)
Sent message. "what was the price og gbpusd on last friday" (neither did Bing/ChatGPT initially)
Searching for: price of gbpusd last friday (Bing/ChatGPT self corrected the typo)
Bing/ChatGPT answered: "I’m sorry but I couldn’t find the price of GBP/USD on last Friday. However, I can tell you that as of today, the exchange rate of GBP/USD is 1.37681." (Really, I'll have some of that! :) )
Me, having noticed the typo and thinking maybe that caused the wrong price quote: "what was the price of gbpusd on last friday"
Bing/ChatGPT: "I’m sorry for the confusion. According to the data I found, the exchange rate of GBP/USD on Friday 7th July 2023 was 1.268321."

Interesting conversation, good that it self corrected but not so good that it 'got it wrong' first time round. Does that mean we cannot rely on it's answers at first flush?

I initially asked the first question to test it's 'uptodateness' as Bing claims it provides real time search info to ChatGPT for up to date replies whereas standalone ChatGPT is (still?) limited to pre Sept 2021 data.

Stop Press: It's just done it again! ...

Me: "what is the current gbpusd exchange rate"
Sent message. "what is the current gbpusd exchange rate"
Bing/ChatGPT: "As of now, the exchange rate of GBP/USD is 1.3768".

I thought I would try OpenAI for the current exchange rate and got the "I'm sorry, ... I don't have real-time data or internet access to provide you with the current exchange rates.". So that rather demonstrates the Bing version is perhaps more up to date, albeit it got it wrong! :o

Edit: I thought I would give Google Bard a try so I asked it the same question re the current exchange rate and it correctly replied "The current GBPUSD exchange rate as of July 10, 2023, 01:47:26 PST is 1.2834." So I tried it on the (corrected) 'last friday' question and it again got it right responding "The price of GBPUSD on last Friday, July 7, 2023, was 1.2838". Then just for cheeky completeness I asked it the original question with the typo and it still got it right. :)

Not sure what to glean from this one-off sampling but Google Bard was the winner.
User avatar
Realrocknrolla
Posts: 1910
Joined: Fri Jun 05, 2020 7:15 pm

https://www.youtube.com/watch?v=i0tzkuq ... 7pWMlyV4p-

1. **Introduction (00:00 - 01:08)**: The presenter begins by acknowledging that Betfair trading can seem complicated due to the many factors to consider, such as predicting the direction of the market, deciding when to open a trade, and whether to back or lay first. He then introduces the strategy he will discuss, which he claims is simple to understand and execute, and doesn't require guessing the market direction.

2. **Setting Up the Ladder Trading Interface (01:08 - 03:37)**: The presenter demonstrates how to set up the ladder trading interface on Bet Angel to facilitate the strategy. He adds the 'estimated position in queue' feature, changes the traded volume to a percentage, and adjusts the back and lay columns to show every cell with money available. This setup helps to highlight gaps in the market and identify areas with small amounts of money.

3. **Executing the Strategy (03:37 - 07:22)**: The presenter puts the strategy into action in a live market. He places a closing order at a price point where there's very little money already in the market, which puts him near the front of the queue. As the market price moves towards his closing order, he places an opening trade. The aim is to get both the opening and closing trades matched quickly, minimizing risk. He repeats this process a few times, each time making a profit.

4. **Summary (07:22 - End)**: The presenter concludes by reiterating the simplicity and effectiveness of the strategy. He acknowledges that there's still a risk involved and that traders may occasionally get into the market too early and see the price reverse. However, he assures viewers that with practice, they can improve their execution of this strategy. He also hints at additional features in Bet Angel that can further facilitate this strategy, promising to cover them in a future video.

The strategy discussed in the video is essentially a form of queue trading. It involves placing a closing order at a price point with little money, which puts the trader near the front of the queue. As the market price moves towards this closing order, the trader places an opening trade. The aim is to get both trades matched quickly, minimizing risk. This strategy does not require predicting the market direction, making it relatively simple to execute.

"can you be a little bit more detailed please"

Absolutely, let's delve deeper into the video:

1. **Introduction (00:00 - 01:08)**: The presenter, Peter Webb, begins by acknowledging the complexity of Betfair trading. He mentions the numerous factors that a trader has to consider, such as predicting the direction of the market, deciding the right time to open a trade, and whether to back or lay first. He then introduces the trading strategy he will discuss in the video, which he claims is simple to understand and execute, and doesn't require guessing the market direction.

2. **Setting Up the Ladder Trading Interface (01:08 - 03:37)**: Peter demonstrates how to set up the ladder trading interface on Bet Angel to facilitate the strategy. He does the following:

- Adds the 'estimated position in queue' feature: This feature shows the trader's position in the queue for a particular price point. It helps the trader understand how close they are to having their order matched.

- Changes the traded volume to a percentage: This allows the trader to see the volume of trades as a percentage, which can provide a clearer view of market activity.

- Adjusts the back and lay columns to show every cell with money available: This highlights the price points where there are pending orders in the market. It helps the trader identify gaps in the market and areas with small amounts of money.

This setup helps to highlight gaps in the market and identify areas with small amounts of money, which are key to executing the strategy.

3. **Executing the Strategy (03:37 - 07:22)**: Peter puts the strategy into action in a live market. Here's a step-by-step breakdown of his actions:

- He identifies a price point where there's very little money already in the market and places a closing order there. This puts him near the front of the queue, increasing the likelihood of his order being matched quickly.

- As the market price moves towards his closing order, he places an opening trade. The aim is to get both the opening and closing trades matched quickly, which would complete a successful trade and minimize risk.

- He repeats this process a few times, each time making a profit. He does this without predicting the market direction, simply reacting to the market's movement towards his closing order.

4. **Summary (07:22 - End)**: Peter concludes by reiterating the simplicity and effectiveness of the strategy. He acknowledges that there's still a risk involved and that traders may occasionally get into the market too early and see the price reverse. However, he assures viewers that with practice, they can improve their execution of this strategy. He also hints at additional features in Bet Angel that can further facilitate this strategy, promising to cover them in a future video.

In essence, the strategy discussed in the video is a form of queue trading. It involves identifying gaps in the market, placing a closing order at a price point with little money, and then placing an opening trade as the market price moves towards the closing order. The aim is to get both trades matched quickly, minimizing risk. This strategy does not require predicting the market direction, making it relatively simple to execute.
User avatar
danielee
Posts: 1
Joined: Sat Apr 01, 2023 9:09 pm

I tried writing e-mail messages using chptGPT, but I didn't like it. It writes too primitive I think. Or maybe I just didn't understand how to use it
User avatar
Trader724
Posts: 576
Joined: Fri Dec 27, 2019 11:22 pm

An NLP is only as good as the prompter.
User avatar
ShaunWhite
Posts: 10374
Joined: Sat Sep 03, 2016 3:42 am

Trader724 wrote:
Thu Jul 13, 2023 11:55 pm
An NLP is only as good as the prompter.
There is a compelling argument for enhancing sentence construction and vocabulary within the school curriculum. If people aren't unable to articulate their thoughts concisely, they won't fully benefit from it.

It's like y'know garbage in and man dem get it back yeah, you get me init.
Post Reply

Return to “AI, Machine Learning and Generative AI”