I see, I did a double take and realized people call this "cumulative delta". As mentioned, I don't use such techniques; it's not up to me to conclude if there's any merit because it just means I haven't spent a lot of time with it.



Well, in finance, the opposite of being "right" is being "random". "Poor or even negative sometimes" sounds to me like a random distribution of outcomes.



Just completely different models; better ways to transform that data into useful insights, and more accurate ways to sign the trades.

One issue is that you have to contend with is the cause and effect when you're counting volume in increments and using that to forecast that it precedes any impending volatility (or spikes or trends or whatever the practitioners call it in this space). Naturally, sizes of the individual transactions decrease as the order book depth thins, and any attempt to make a forecast using individual transaction sizes without removing this baseline phenomenon is going to be ineffective because your signal is dominated by the baseline phenomenon.

Also, the ironic thing is if you actually improve the signing of the trades, you'll often find that the performance actually decreases.

They are making 20% to 40% with their ''randomness'' per annum.What are the standard for being ''right'' then?If we assume that market offers up to 1% per day(just a single underlying),being ''right'' may mean 20% per month(a single underlying),at least.

@artemiso,

i was trying to address my question to how the HFT industry dismantled volume/price analysis,but your reply was somewhat vague.

I don't think the industry has dismantled anyone's trading strategies; the FIA maintains a literature review of academic studies on this topic from unbiased researchers (e.g. https://www.futuresindustry.org/ptg/downloads/HFT_Literature_Review_0913.pdf) and in fact most studies suggest that we improve price discovery and reduce volatility and trading costs, all of which I understand that chart-based traders have always wanted.

The amount of trading volume that chart-based traders contribute to the market is very small, so it makes no sense why our industry would pick on chart-traders of all participants. Moreover, by saying that we 'dismantled' volume/price analysis, you're implying that we compete on the same strategies. I disagree on that for reasons I've described before - that I think that volume/price analysis in the "cumulative delta" sense is misguided.

You`ll never know if they are biased or unbiased.Moreso,when the scholars used terms such as ''seems/seemed'' it doesn`t look like scientific research,it seems

It does say in the paper that the HFTs trade alongside with the other participants,though.But the buzzword is 'frontrunnig'(meaning they in early and out early),so it surely affects the chart trader`s V/P analysis.

I also belive that HFT has nothing to do with narrowing spreads,lowering trading costs and volatility.It`s more like due to industry expansion and competition,availability of technologies.

It's very easy for us to see that we're the most critical factor in reducing spreads for market participants when we're making up 10% of the trading volume in a single product at times or being 1-5 of the only market makers in some FX flow.

Availability of technology cannot have been the explaining factor for the reduced spreads. In fact a few of the trading venues with the fastest matching engines and market data processors are actually also the ones with the widest spreads because they've failed to attract competitive electronic market makers. One comparison that comes to mind is Canada (TSX) vs Tokyo (TSE) - their matching engines were neck-to-neck but the progression in spread size is drastically different if you look around the dates of Tokyo's introduction of arrownet vs Canada's regulatory attempts to deter substantial order modifications.

Frontrunning is an activity where a broker or fiduciary executes for his own account ahead of his client, using information about his client's orders. Electronic market makers generally have no brokerage clients, and therefore it is impossible for them to frontrun.

Moreover, by saying that we 'dismantled' volume/price analysis, you're implying that we compete on the same strategies. I disagree on that for reasons I've described before - that I think that volume/price analysis in the "cumulative delta" sense is misguided.

Regarding the latter, I haven't been using models that attempt to determine some underlying function that maps from the prices to volumes;