How do HFTs screw small hidden limit orders posted inside the spread?

[topdrudge]

Was your strategy a profitable intra-day strategy? Were the stocks impacted by news action?
However, my case is different. I will accumulate throughout the day in stocks impacted by no news action and keep the positions over many days. There is no urgency. I am just looking for a cheaper entry and exit.

At the time I was talking about, my strategy was a lot like yours.

Orders get filled at other exchanges at better prices, because the hidden orders do not interact with orders at other exchanges and because the lit orders have higher priority at the same price leading to what you describe as “huge numbers of trade-throughs, along with excessive numbers of trade-ats where I got nothing.”

The trading through or at your limit price = slower speed of execution imply that some of the inherent profit of your strategy was left on the table. Your P&L may then stay the same.

Right.

I will get traded through and at my limit on other exchanges. But at the exchange were my order resides, I assume the game will be less rigged against me. I noted your three points above and I also experience the pinging where HFTs try to expose my hidden orders, but still I feel less molested compared to posting regular limit orders.

Regarding "I assume the game will be less rigged against me" and "still I feel less molested" ... right, that is the seductive thing about using hiddens ... I felt the same way back then, and I still use hiddens occasionally when I just want to avoid the pain of feeling like a prize sucker competing with the HFTs ... but if you keep careful track of your metrics I think you will reach the same conclusion as I did: It's just psychology; you're not really gaining any trading advantage.

When I use regular limit orders, I immediately enter an arms race against other bids that increases the bid - often to the point that I have to pay the full spread.

Sure, we all experience that, every day, all day ... but the fairest comparison is not between hiddens and penny-jumping; rather, it's between hiddens and joining. Anyway, if you want to remove the psychology and get some real metrics you should compare and contrast all three for your strategy: joining, penny-jumping and hiddens.

 

[Bob111]

Beyond that ... you are trying to grab the same pennies that HFT is going for, and you're not likely to win at that game.

yeah..been trading stocks for a long time and have to put my existing trading systems on halt for this particular reason mentioned by you above. i know . from what i have seen-my chances to win are pretty slim ,but you know..
“Every accomplishment starts with the decision to try.”


i would add-or failure

 

[Daal]

I did a big experiment with hidden orders a year or two ago, back when IB first coded them up correctly. For several months I used a large proportion of hiddens inside the NBBO, along with non-hiddens and compared the results, both with recent performance of my trading system with no hiddens, and comparing hidden and non-hidden performance during the experimental period. Experimental data points were mainly daily p&l and number of executions, so this was not a detail-rich experiment, but it was somewhat controlled. We're talking about on the order of 100,000 hidden orders.

Results: I noticed no significant performance difference (one way or the other) between hiddens and non-hiddens. The trading was a little different at the microstructure level, but overall results were essentially unchanged. If I had to lean one way or the other I would say performance with hiddens was somewhat worse (as you can imagine there were huge numbers of trade-throughs, along with excessive numbers of trade-ats where I got nothing).

My results are stale and coarse ... would love to hear from other elitetraders with more timely comments, or more specific data.

See my post above: My personal opinion is that "hiddens" are not hidden at all, unless it so happens that for a particular order nobody cares about sniffing you out. If anybody wants to sniff you out, they know all about your pathetic hidden order and are laughing at you.

What is the price range in the stocks you tested? Avg spread size? When you had your trade throughs didn't you find useful to use multiple hidden orders at the most liquid pools instead of just one?

 

[topdrudge]

What is the price range in the stocks you tested? Avg spread size? When you had your trade throughs didn't you find useful to use multiple hidden orders at the most liquid pools instead of just one?

We're talking about all kinds of price ranges and other market behavior since the experiment stretched over a period of months and involved hundreds of symbols ... average spreads for the most part were at least a few cents (lower-volume symbols); in my opinion it would be a completely different (but still interesting) experiment if using high-volume large caps with <$0.01 spreads.

Sure, multiple hiddens at different venues would reduce trade-throughs ... but at the obvious cost of getting swept all the time for far more shares than I want ... not my style of trading.

 

[Daal]

We're talking about all kinds of price ranges and other market behavior since the experiment stretched over a period of months and involved hundreds of symbols ... average spreads for the most part were at least a few cents (lower-volume symbols); in my opinion it would be a completely different (but still interesting) experiment if using high-volume large caps with <$0.01 spreads.

Sure, multiple hiddens at different venues would reduce trade-throughs ... but at the obvious cost of getting swept all the time for far more shares than I want ... not my style of trading.

If you test in the high priced larger spread stocks (like AAPL or TSLA), I believe you will notice more improvement.

Also, you simply tested 100 shares in one ECN or exchange? If you have 1000 shares that you want to to buy or sell and break down in 3 hidden orders, there might be additional improvements that you didn't see in your test

 

[Grandluxe]

They sniff you out with a fucking odd lot. But it doesn't matter now.

The problem right now is there aren't enough opportunities. You cannot get enough stock at the price you want. And if you do, guess what? It runs against you.

 

[topdrudge]

If you test in the high priced larger spread stocks (like AAPL or TSLA), I believe you will notice more improvement.

Also, you simply tested 100 shares in one ECN or exchange? If you have 1000 shares that you want to to buy or sell and break down in 3 hidden orders, there might be additional improvements that you didn't see in your test

Points well taken Daal, and thanks for the post ... but I would remind all readers that in my case we're talking about a trading program that no longer exists and observations that are 2 years old ... the main reason I started posting to this thread was not to give anything useful, but rather to smoke out some posters who could provide fresh, new observations supported by facts and careful analysis. For example, take Daal's first statement quoted above ... this is a great question that has answers which can be discovered relatively easily ... can anyone comment intelligently on it?

 

[Maverick2608]

Anyway, if you want to remove the psychology and get some real metrics you should compare and contrast all three for your strategy: joining, penny-jumping and hiddens.

May I ask what you prefer between joining, penny-jumping and hidden?

I asked the same question to an expert on market microstructure who has written several articles in CFA Magazine: Can you avoid HFT sub penny front running by posting small hidden limit orders inside the spread?

He seems to agree with the results of your test. This was his answer:
"Unfortunately, this will not help. The SEC requirement for HFT internalizers is that they match or beat the NBBO. That is the "displayed" quotation.

For example, if the NBBO on stock XYZ is 25.50 to 25.60, under your scenario you could place a hidden bid at 25.51. The internalizer can still transact at 25.5001 (through your 25.51 bid) because your bid is not part of the NBBO (since it is hidden). The only thing hidden bids can help with is the algorithmic penny-jumping, where an algo steps ahead of your displayed quotation by a penny in the displayed market. Hidden orders do nothing to stop the internalizers.

This is something the SEC should really investigate further. Market orders being transacted by internalizers do not have the chance to interact with this hidden order flow. With so much hidden liquidity on the exchanges + hidden liquidity in various dark pools, the retail market order is in many cases not getting the best available price. Everybody seems to lose here, except the internalizer. "

 

[topdrudge]

These days my most prevalent pattern is to first try penny-jumping (never hurts to try, and even if you are joined instantly you still - hopefully - have some priority), then if I get jumped instantly, I cancel and join instead. I only use hiddens in very special circumstances, like joining small size with larger size.

Your expert is talking about the BIG problem: broker-dealer internalization and dark pools ... that's a completely separate subject from this thread. Comparing the damaging effects (for lower volume stocks) of internalization with what predators do on the lit exchanges, internalization ranks as at least 10 times worse in terms of killing opportunities for profit.

 

[Bob111]

They sniff you out with a fucking odd lot. But it doesn't matter now.

The problem right now is there aren't enough opportunities. You cannot get enough stock at the price you want. And if you do, guess what? It runs against you.

yep. i was working on one really short term( few min max) system that i described earlier in some other thread. everything is fine ,good PnL on backtests, even if i put entry at midpoint between bid \ask,exit on stop at worst-ie sell at bid,buy cover at ask,lot of trades etc...until i added one simple condition-count shares and transaction prices that are at my entry price or better for 5 seconds,after a signal. if there is at least 300 shares traded at my price or better-i'm in.
and everything fell apart(speaking about importance of backtesting discussed in some other thread).