What's the definition of a good fill and how do you know you aren't getting good fills on shorts? Negative PnL proves nothing unless you have a massive sample size. If you suspect there is an issue where sshort orders have more latency than you expect, you should be able to prove this by cross referencing your executions with the tape.
Well, not any individual fill could be necessarily said to be good or bad in most cases; but at the end of the day I have my net prices calculated for buys, sells and sell_shorts, and its almost always the case that avgBuyPrice>avgSellPrice but avgShortPrice < avgBuyPrice . Perhaps thats just happenstance though, because when I went long in the mornin my max inventory size so that i could both buy and sell at the same time without going short, it still came up negative pnl at the end of the day . and im still having problems with double-fills, getting filled before the cancel request arrives at the exchange, so now I gotta add in logic to wait until the order is canceled before inserting a new one. Maybe I should just try using the exchange's primary peg orders rather than trying to do it myself. This also explains it: http://www.nasdaqtrader.com/TraderNews.aspx?id=ETU2012-11 How are Primary Peg Orders with an offset amount changing? UPDATE:NASDAQ, BX and PSX will change the treatment of Primary Peg Orders with an Offset to be non-displayed. This change is being postponed to allow firms more time to prepare. A new production date will be announced in the near future. Why is this changing? The display of Primary Pegs with an offset can potentially result in excessive messaging when multiple venues display pegged non-marketable orders. In these scenarios, it is possible for the Primary Peg orders on each venue to react to and change in relation to each other, resulting in excessive messaging and "quote flickering". By changing the treatment of Primary Peg Orders with an Offset to be non-displayed, NASDAQ, BX and PSX will prevent this feedback loop, adding to system stability and improving market quality. This also looks like it will help https://www.tradersmagazine.com/am/the-iex-d-limit-proposal-its-goodbut-what-if-its-too-good/
You need to have a way of measuring the performance of your systems that can objectively tell you if things are working as expected or not. The fact that you are looking at one day of pnl being negative and thinking this could mean anything whatsoever is concerning as it sounds like you may not have this figured out. I'm curious why you think you can make money doing HFT market making in the first place. Do you have an edge for picking out high expectancy trades or are you just quoting everything? How many trades/day are you doing in how many symbols?
This also explains it: Because I implemented a strategy described in a PhD thesis and simulated it with realistic fees on a multi agent simulator and got results that matched the thesis. It quotes either at the BBO or improved by one tick on each side in order to have a mean-zero inventory . The problem im having is getting picked off when trying to keep the order pegged/in the queue. Im doubting it because the negative pnl has been about 14 days now.. fixing bugs one by one and now it's still getting picked off.. therefore it seems reasonable to expect using IEX peg orders rather than trying to roll my own order pegging code would work. I dont need to create a histogram of order ack latencies to see im being picked off. Its obvious in the blotter which has nanosecond resolution. Iex peg orders are fair and retain their insertion order as the bbo moves which is not true of the nasdaq which can't guarantee order of reinsertion when it moves , so im going to give that a shot. Its challenging to make a perfect algo not react to its own updates..in the simulator I used zero intelligence market order submitters who arrived randomly, not simulating predatory sniping behavior that iex designed their crumbling quote indicator to avoid https://sbfc.sydney.edu.au/program/papers/P081_Named.pdf
So let me get this straight, you simulated a fake market where all other agents trade with "zero intelligence" and this is your basis for thinking you can make money in the real market? Hate to break it to you but the real market is extremely competitive. You need to have an edge when you trade. Using limit orders at BBO isn't an edge. In the first place if your orders are on the exchange you are barely interacting with dumb money at all. Internalizers own that game now.
Yep, and/or they sell flow to HFT internalizers. Doesn't mean you can't still make good money providing liquidity on the exchanges. But the days where you could print money simply buying the bid/selling the ask are long gone.
I figured they were. I read a study that showed just constantly quoting at bid and ask without regard to inventory leads to ruin no doubt. I have read other strategies where it can be done, if you can get your orders to be filled in such a way as to be balanced on average, and the volatility and adverse selection are not so great as to negate the profit from making the spread. In any case, im going to try quoting on iex before I throw in the towel. I've put in too much work to give up easily
Is speed the primary factor now since everybody more or less uses the same factors for prediction in hft mkt making
Take this with a grain of salt because I have never worked inside an HFT. I've read a fair bit about them though and spent quite a bit of time researching market structure. It really depends what we are talking about when we say HFT since the term means a lot of different things to different people. If we are talking about "classical" "market making" HFT which is basically rebates and making the spread: My understanding is these edges have been competed down to the point where scale becomes everything and barriers to entry are now almost impossibly high. So I would say scale of the operation is the primary factor, and top-tier speed is simply a given. For example if the average profit on such a "classical" trade is under 0.001/share before fees, then to be profitable your fees need to be multiples lower than the cheapest brokers. You basically have to be your own broker-dealer and get volume discounts from the exchanges. To do enough volume you need to trade every stock on every exchange. You need to be colocated in every exchange and consuming their raw feed because the SIP is too slow. Such firms typically also have other lines of business like PFOF internalizing or market making other assets that can subsidize their costs during lean times so they can stay in business and cash in when volatility appears. Over time these firms are naturally consolidating towards a cartel or monopoly simply due to the market structure. Basically these edges are inaccessible in their pure form if you don't have the right scale. The only solution for smaller traders is to find niches within the market that have higher profitability. Instead of taking the other side of every trade, only take trades where your average profit is significant for you after fees. For retail, this is typically going to be 0.01/share and above. The trick is being able to identify those niches.