AI Models

[guspolinkski]

Anyone using AI to build sports models?

 

[Tanko]

You have to be very careful using AI for this purpose. It makes a lot of mistakes.

The prompt you use must be relatively specific/detailed if you want to try and get good data. Even then it has flaws.

I would say AI is useful in providing ideas for developing models but as far as using it to create models for you, it's not there yet.

Just my opinion.

 

[guspolinkski]

$200 month subscription gets you 5.2 pro. You can feed it datasets, do heavy reasoning. I would feed I it the input data and target predictors, not rely on AI to gather data. Maybe not build an entire model from scratch, but help you analyze data, select features, develop a plan for experiments. Help you build models and design a pipeline. 5.2 pro and codex high can help you develop a plan, build the software, do the testing, etc

Tackle more sports, more predictors, do all the coding, optimize code you’ve already written

Even $20 subscription gets you codex with rate limits, might start with that

 

[KVB]

AI doesn't know shit when it comes to sports.

I was so worried about it overtaking my work, but I'm not so worried now, based on what I've seen.

That day will probably come, though.

 

[ktollinche]

garbage in, garbage out.

If the creator is not already a winning gambler, I doubt those things will help much

 

[maltedhopsfrenzy]

What a disappointing thread!
Came in here looking for this:

 

[Tanko]

What a disappointing thread!
Came in here looking for this:

View attachment 87006

NICE! Malted knows how to use AI to its fullest extent.

 

[jjgold]

If you do not have access to APIs and large databases AI is worthless

 

[guspolinkski]

AI doesn't know shit when it comes to sports.

I was so worried about it overtaking my work, but I'm not so worried now, based on what I've seen.

That day will probably come, though.

Models are rapidly improving. Not sure what you’ve seen, how it was being used, how recent, and if you’ve seen what 5.2 pro can do.

AI might not know shit about sports, but I’d think it knows about modeling, data analysis, and setting up experiments. It can learn if you guide it. It certainly knows about coding. As an assistant to someone with a background in software, modeling, and sports betting, I think it has tremendous potential to bring immediate value.

You have credibility, but there are some things you have to do your own research and try for yourself. I’m not ready to count out AI yet and comfortable risking a failed attempt.

 

[guspolinkski]

If you do not have access to APIs and large databases AI is worthless

Just because someone hasn’t found value with sports betting from AI, doesn’t mean it’s impossible. Maybe others can use it differently and find value.

 

[KVB]

Models are rapidly improving. Not sure what you’ve seen, how it was being used, how recent, and if you’ve seen what 5.2 pro can do.

AI might not know shit about sports, but I’d think it knows about modeling, data analysis, and setting up experiments. It can learn if you guide it. It certainly knows about coding. As an assistant to someone with a background in software, modeling, and sports betting, I think it has tremendous potential to bring immediate value.

You have credibility, but there are some things you have to do your own research and try for yourself. I’m not ready to count out AI yet and comfortable risking a failed attempt.

Oh I'm definitely not counting it out.

I agree with everything you say.

It's just not there yet.

Like I said...

That day will probably come, though.

 

[guspolinkski]

Oh I'm definitely not counting it out.

I agree with everything you say.

It's just not there yet.

Like I said...

I meant I’m not yet ready to count it out in the present. Maybe for a seasoned bettor like you with mature models who is already successful, it would be less impactful at the moment. But for someone still in the development phase, not using AI seems like a mistake. Especially for coding. Maybe I’ll feed it the KVB’s High transcripts.

 

[ktollinche]

yeah but is it gonna come up with something that beats -110?

People have been data mining the crap about sports for like 40 years now.
(and the data set is not that huge)

AI doesnt do revolutionary (yet), it just speeds your work flow.
But what if there is nothing to 'speed up'

 

[KVB]

Don't even play into -110.

No need to.

I meant I’m not yet ready to count it out in the present. Maybe for a seasoned bettor like you with mature models who is already successful, it would be less impactful at the moment. But for someone still in the development phase, not using AI seems like a mistake. Especially for coding. Maybe I’ll feed it the KVB’s High transcripts.

100%

I'm thinking of using AI for coding an interface for me to work from but I need help.

Turtledoves, check in pal

 

[BobbyFK]

I think AI will get smarter as the years go on, but I don’t think it’s ready yet to build a model that I would trust

 

[phillyflyers]

What about this one?

Introducing Perplexity Computer​

Frontier AI models are getting smarter. The best are becoming so capable that the products built around them are a bottleneck for showing their true potential. And as every new model specializes more, only Perplexity can unify all of the top capabilities into a single system for the most powerful use of AI yet.

Today we're introducing Perplexity Computer.

Perplexity Computer unifies every current AI capability into a single system. Computer is a general-purpose digital worker that operates the same interfaces you do. It is the next evolution of AI. Chat interfaces have answers, while agents can do tasks. Perplexity Computer is a system that creates and executes entire workflows, capable of running for hours or even months.

How it works​

Perplexity Computer operates the software stack just like a human co-worker would: by using it. Computer reasons, delegates, searches, builds, remembers, codes, and delivers.

Start by describing an outcome. Perplexity Computer breaks it into tasks and subtasks, creating sub-agents for execution. The sub-agents might do web research, document generation, data processing, or API calls to your connected services. A document is drafted by one agent while another gathers the data it needs.

The coordination is automatic, and the work is asynchronous. You can focus on other things, or run dozens of Perplexity Computers in parallel.

When Computer runs into a problem, it creates sub-agents to solve it. It can find API keys, research supplemental information, code apps if necessary, and check in if it truly needs you. Every task runs in an isolated compute environment with access to a real filesystem, a real browser, and real tool integrations.

The result is a safe harness for powerful AI, universally accessible without localized setup, and delivering the best work of every current top AI model.

A natural evolution​

Perplexity’s mission is to power the world's curiosity. From the beginning, we knew this goal would require highly accurate AI because, for people with questions that matter, knowledge is an essential edge. They need answers they can depend on, to ask the next question.

We also knew that questions are everywhere for curious people, and some answers require actions. So we built Comet, the world's first truly AI-native browser, and the personal AI agent, Comet Assistant. On these foundations we’ve been building the most accurate AI, advancing Perplexity deep research to be the industry best, incorporating persistent memory, and tasks, and more.

Meanwhile, everything Perplexity builds has always been model-agnostic. Choice and control are essential for better outcomes. Now, as models advance, it’s clear the most powerful AI system isn’t built on any single model or family, it’s the one that can orchestrate all of them.

Intelligent multi-model orchestration​

As models become more powerful, multi-model orchestration makes a versatile AI harness the most powerful one.

Conventional wisdom has said AI models are commoditizing but, we’re actually seeing the opposite: models are specializing. Each frontier model excels at different kinds of work, so a full workflow must have access to them all and deploy them intelligently.

As of this writing, Perplexity Computer runs Opus 4.6 for its core reasoning engine and orchestrates sub-agents with the best models for specific tasks: Gemini for deep research (creating sub-agents), Nano Banana for images, Veo 3.1 for video, Grok for speed in lightweight tasks, and ChatGPT 5.2 for long-context recall and wide search.

Perplexity’s model agnostic harness allows these to change as models advance. Computer also allows you to choose specific models for specific subtasks. As token budgets become a more common consideration in different roles, users must have choice and control over their best use.

AI is now the computer​

In 1757, mathematician Alexis Clairaut employed two "computers,” the title for apprentices at the time, to refine Edmond Halley's prediction about comets. Working day and night for months, the trio split their work and successfully predicted the perihelion of Halley's Comet within two days of accuracy.

The word “computer” has evolved since then, but it’s always meant the same thing: autonomous division of complex work, with accuracy as a central underlying necessity.

The best AI models in the world are already capable of extraordinary work. What they’ve lacked is a system that can orchestrate capability across tasks, tools, and time. Until today. Perplexity Computer is here and available to Perplexity Max subscribers. It will also be available to Enterprise Max users soon.

 

[jjgold]

Don't even play into -110.

No need to.



100%

I'm thinking of using AI for coding an interface for me to work from but I need help.

Turtledoves, check in pal

You don’t need help with ai

I built 5 iOS apps in store

Can get data any website formatted perfectly with some tricky node js code

Can build any trading App If have the proper api credentials

Can do anything on vps server

Going to program and old school pager for fun to give me alerts like the old days from apis

Can build desktop apps

Can build line screens , quote boards

Can Program
Any m5 portable device

 

[Tanko]

You don’t need help with ai

I built 5 iOS apps in store

Can get data any website formatted perfectly with some tricky node js code

Can build any trading App If have the proper api credentials

Can do anything on vps server

Going to program and old school pager for fun to give me alerts like the old days from apis

Can build desktop apps

Can build line screens , quote boards

Can Program
Any m5 portable device

JJ, Build me an app that autobuys/sells large cap stocks with regular dividends, a P/E < 15 and risk profile of 3 or less, and projected to be hit new highs in the next 12 months.

I'd like to be able to add triggers to stock buy/sell executions based on recent trading.

Thx

 

[KVB]

You don’t need help with ai

I built 5 iOS apps in store

Can get data any website formatted perfectly with some tricky node js code

Can build any trading App If have the proper api credentials

Can do anything on vps server

Going to program and old school pager for fun to give me alerts like the old days from apis

Can build desktop apps

Can build line screens , quote boards

Can Program
Any m5 portable device

I could probably build the interface, but there is no way to have the computer trade like I do.

It's a high performance sport in a fluid situation.

I have to have my hand on the trigger.

 

[jjgold]

Hard to beat auto traders
Same on Wall Street impossible