why Backtetsing stocks difficult vs backtesting Futures


Coming from backtesting/optimising Futures for few years using Ninja tarder and Multicharts

which I do the same instrument every day, every week all year, so backtesting is quite simple here from my experience

in the case of stocks, i struggle to backtest, as symbols are NOT the same everytime

say, for exmaple
This week, I run a scan, I get 20 symbols
Next week, I run a scan, I get 25 symbols

what period (days) would I backtest the 1st set of 20 symbols and
what period (days) would I backtest the 2nd set of 25 symbols

I am sure these are traded once and not repeated
and what software for stocks, can we backtest like this?

and how can I backtest a week past or a month earlier for a scan thats developed a setup now,
example 10EMA/30EMA

What you're saying doesn't make sense, or atleast I don't understand it.

Are you scanning a basket of stocks, and then running more specific backtest on each individual underline?

I'm not sure how you could expect symbols to be the same every-time you run a scan, each stock can act like an independent data series. So the only way you'll get the same results is if you're curve fitting.

ok, I am coming from Futures that trades everyday against stocks that only trades when set of condition meets, so once a month or less.
let me clarify
I have a scan setup to run on entire Nasdaq, so come up with a set of symbols and I have this symbols to test my conditions to see if they have an entry
1st issue. entire Nasdaq or NYSE, is quite high # of symbols, about 3000. not sure NT can handle it. havent done it

The Naz and the NYSE are very different. But aside from that, if you're getting too many hits, narrow your universe, such as by float, both too much and too little. That may decrease the number of hits dramatically. You may also narrow it according to those who actually earned something last year and those who didn't.

thanks for your input. my strategy doesn't have criteria to filter and scan Fundamental screening, but i dont understand " by float" pl explain

but you got my objective right. I do want to filter and less trades/symbol in a year

A universe of a hundred highly-desirable stocks is plenty. You can review these in a few minutes if you know what you're looking for. Thousands is excessive.

Exactly, there's no point scanning multiple stock indices when many stocks could have completely different characteristics eg volatility/volume/ liquidity- you could expect the data set on alot of stocks to be independent of one another. So you need to know what you're looking for, and at-least have a basic criteria that filters out the junk to a universe that is tradeable and shares more of the same characteristics. Once you've optimized it then be sure to out of sample test it.