First AI based Screen

judgetrade · September 27, 2024, 6:11pm

Not spectacular, but making progress

ZGWZ · September 27, 2024, 6:14pm

What is your universe?

judgetrade · September 27, 2024, 9:54pm

build it on my own, basically as small cap universe...

judgetrade · September 27, 2024, 9:59pm

Here is a system based on the AI Factor 75% + 25% the Sentiment ranking system

It picks different stuff than my other models, nice for diversification...

judgetrade · September 27, 2024, 11:30pm

Here the first Mid Cap - Big Cap model --> took almost everything out of Estimates and Actuals and put it in the bucket.

Same ML Model as with in the small caps...

Ranking Model uses Sentiment + bigger Market caps get better ranking + relative strength + the AI Factor.

Gee, start to like that stuff...

Andreas

ZGWZ · September 28, 2024, 5:55am

What is your volume filter?

judgetrade · September 28, 2024, 8:34am

on the small cap universe AvgDailyTot(20) < 5000000

ZGWZ · September 28, 2024, 9:03am

Do you mean the P123 sentiment system or your own system?

ZGWZ · September 28, 2024, 9:29am

So what is your minimum liquidity requirement? I see the "min vol" in your universe's name.

judgetrade · September 29, 2024, 10:28pm

none --> but the ranking system favors stocks with EPS Estimations --> and they usually have a bit higher volume (Tradable for me), also standard P123 slippage formula is used in the strategy

Here are the holdings of today (for checking the volume...)

AIRI
ALLT
CRNT
CSBR
EEX
ELMD
FSI
GTEC
IPW
MG
NGS
NGVC
NWPX
QUAD
RCMT
RDCM
SHIP
SLNG
SMWB

judgetrade · September 29, 2024, 10:30pm

yes P123 sentiment...

judgetrade · October 3, 2024, 4:13pm

same model trained on higher liquid universe

judgetrade · October 3, 2024, 4:18pm

now here the old model and I added the min liqu. requirement to the buy rule

looks much better:

ZGWZ · October 3, 2024, 4:24pm

Looks like I need to try out the idea of adding liquidity requirements to the buy rules instead of adding them to the universe rules.

judgetrade · October 3, 2024, 4:27pm

yes, I think it is better to train without a minimun liquidity but then put the min liqudity rule in the buy rule...

The data in the small space is sparce enough, so not limiting the universe I think is a good idea, the more data the ML finds the better...

judgetrade · October 3, 2024, 4:31pm

original model with MedianDailyTot(126) > 100000 in the buy rule (it degrades, but cap curve still looks good...)

test_user · October 4, 2024, 4:28pm

Thanks for suggesting this – I have my own method for optimizing ranking systems, and increasing the size of the universe by including the smallest nanocaps really helped!