Comment by chumzygood | Hacker News Reader

Comment by chumzygood | Hacker News Reader

I run a small experiment on this: 29 paper-trading accounts on real US stock prices, $100k each, since July 27. Four AI models (ChatGPT, Claude, Grok, Gemini) each write a trading rulebook and rewrite it every day from their own results. One account is a fixed rulebook that no AI ever touches, as a control.

Result so far (paper, 34 trading days): the no-AI control is +13.0%, the S&P 500 is +3.4%, and 25 of the 28 AI accounts are below the control. The best single account is a Grok-written "patience" book at +40%, which I treat as one lucky account in a choppy market, not a finding. At the trade level the AIs and the control look the same: 3,212 closed positions, median +0.06%, median hold about 2 hours. They trade a lot and mostly go nowhere.

Everything is public, including the losses and the retired strategies: https://aitradingcompetition.com/which-ai-is-winning.html and the full trade file as CSV at https://github.com/ckamelhar-collab/ai-trading-arena-data. Paper money only, not advice, nothing for sale on those pages.

Famously Norwegian cows “picking” stocks by shitting in marked grid did as well as professional stock brokers and much better than astrologers!

Love this. Stay hungry, stay bullish. :D

But is that before or after fees? Does each fund manager take more or less than the cow's dollar equivalent in pasture?

before fees and milk

there's also the (apocryphal?) tale about a brokerage investigating which of their clients' stock portfolios had the best performance, and discovering the best performing portfolios belonged to the clients who were dead.

https://www.morningstar.com/columns/rekenthaler-report/archi...

None of this really means anything. For one if you test enough strategies one by luck will be killer. Second, past performance != future.

One AI might consist of some convulated strategy of who knows what, but all it takes in the end is for it to say...hold more tech stocks than not, or buying more call options or leveraged positions in a bull market, being more out of the market during a bearish market, etc. But that same strategy will fail miserably as soon as the market conditions change

yeah the alternative test is to arm a toddler with an account and just have it randomly guess what to do. Even it might outperform the best fund managers over a given timeframe but that's just how the markets work. Over the long run, the toddler will go bankrupt.

Don’t let AI write for you. Everyone can tell.

“Not a finding” lol

The most likely outcome from a naive attempt to automate trading is random results.

Wrong:

Stock prices are not random

(in a sense of "really random every day")

Why didn't you let the bots backtest and forward test to pick the mix of best strategies/indicators?

The problem is the bots may have some of the backtest data already in their training.

Why would that matter? It's either a good strategy or it isn't.

Personally I think hobbyists that think they can second guess institutional finance are just fooling themselves.

Ok for example my strategy is buy Moderna. That's it. Backtest that over the last year.

Agreed on 2nd point. Especially with no real hypothesis i.e. just throw AI at it.

Timing is the strategy though. Buying Moderna is either a good idea or a bad idea depending on when you do it.

If you have been trying to buy MRNA in the last 12-18 month, you are an idiot:

The stock price clearly fell for a very long time; only teethgrindlingy I would see one or two entries for "lets-try-it-fingers-crossing"; the price went better in the last couple of month, though no really strong long-entry signal, maybe 2 in the last few month.

This is what I see in TradingView - for this,you shouldnt apply AI.

I'm curious what signals they're trading on. SEC filings? Candlestick voodoo numerology?

> nothing for sale on those pages

"$20/month · or $160/year"

-- https://aitradingcompetition.com/playbook.html#pricing

You have $2.9M just for this? Wow.

paper trading means that no money is used

Thank you, never heard the term before.

Not to be confused with "commercial paper" which is a short term promissory note in real money.

https://www.google.com/search?client=firefox-b-d&q=finance+i...

They mentioned paper trading which means it isn't using actual money.

> paper-trading

come back when you put it real money and have some skin in the game.

I wouldn't trust AI with real money though. (I barely trust myself with it lmao)

There is one. A real Alpaca account has mirrored one of the bots since August, capped at $3,000, with every fill posted: https://aitradingcompetition.com/real.html. It's small on purpose.

Reply on news.ycombinator.com