I recently came to take a look at penny stocks having trawled through Tim Syke's somewhat embarrassing website. I attach a back test which seems to show vast profit.

I think this code originated with Charles Witt here on Quantopian but in any event I nicked it to see if Syke's chart reading could be duplicated algorithmically.

The point here is that nobody has or can take account of the apparently vast bid offer spreads in a penny stock back test. To do that you need fine resolution bid/offer data not minute data. Without bid offer data too often your limit buy order will be filled at the bid and your limit sell order will be filled at the offer. If the spread is anywhere between 50 and 100% then clearly any profit shown would be a nonsense.

I believe Charles and others discovered that when they tried live trading.

That does not mean you can not benefit from penny stocks but my suspicion would be you need to ride on the coat tails of those who are organizing "promotions". Rather than trading algorithmically. But without bid offer data it is difficult to tell. Of course I am not suggesting doing anything illegal or that anyone such as Sykes is doing anything illegal. Provided you are not an insider, there is no illegality in following somebody else's promotions.

Having said all that, it appears that volume expands exponentially and spreads narrow likewise during pumps and dumps - legal or otherwise.
I have seen reports of live trades conducted at tiny percentage profits and losses.

It seems to me therefor that the key to testing more realistically using minute data has to be to trade on high volume only. Potentially by back testing on this basis and using appropriate volume based slippage it may be that more realism can be introduced.