Payment for order flow

[qaz]

I have made a series of counter arguments on Tabb but I would like to post here for those who are not aware of the implications of PFOF.

Simply extrapolate the factors and you will see how bad it gets. Imagine the HFT firms get bigger in the future and 85% of orders go through them. All traders see the Bid and Ask listed on the exchange and it is more or less stagnant. the 15% is “making the market” and that is the quote that the HFT is guaranteeing to improve against.

Is the Bid Ask quote by the 15% a fair market value?


PFOF buyers are paying for information. If they capture a large volume of all trades (especially now with more brokers going zero comm), they can withhold such information from the exchanges. The actual market can now be easily manipulated like penny stocks.

Dont forget that these PFOF buyers can also do internalisation and profit big time at the expense of the trader that thought he got the best price.

As long as the publicly listed market quotes are not indicative of the real supply and demand (ie ineffecient), the entity with the most information gets to profit the most, that is the PFOF buyer.


Another point I would like to bring up is these MM guaranteeing NBBO but not at the mid. If you trade options, always going for the marketable price will wipe you out quick. If I were to route direct to the best exchange, i would be getting filled at mid or just slightly off most of the time. The MM doesn’t fill at that price but much further from mid or maybe at the marketable price, why? Because they pay for PFOF and need to make at least a penny from me.

 

[guru]

Yes, there will be positives and negatives for different people, but does it help that everyone has a choice to pay zero commissions vs paying commissions and routing orders to exchanges (like IB Lite vs IB Pro)?

 

[qaz]

I will need to point out that active traders will lose out the most just like how a wallet has a small hole and coins drop out ever so often. I believe ET is populated with active traders.

For the crowd that buy and hold for months or years, there's minimal impact.

 

[dinn13]

The HFT firms want non-informed/zero alpha flow and take losses from informed flow that comes along with it. To put it another way, they want to pay for trades that aren't correlated with future price movement.

I was speaking to someone senior at one of the HFT firms and they were complaining that the retail designation is too broad. They want to segregate infrequent from frequent retail like is done with options. In the options market if you place more than 390 orders on average per day then you're classified as a professional even if trading in a personal account and not affiliated with any financial firm.

That's likely why for example IB lite doesn't work with TWS or over FIX. Because that flow is more likely to be informed and the HFT firms don't want it. Most if not all of the firms buying retail flow also let broker algo flow ping them because that flow is for the most part similar to retail but they won't let me directly ping their SLP's cause they see my flow as toxic. They track markouts from each client and will shut them off if negative for them. So they'll never get a majority of the flow because they don't want it. They only want the good stuff.

So there is an opportunity here for active traders that actually have alpha. For example let's say you trade through IB tws/fix and are cost+. If you want to buy and the only quotes are on nasdaq and arca then if you take liquidity you'll pay the exchange fees of 30 mils + commissions and you'll cause significant market impact by directly causing the price to move since the price level is removed. On the other hand if you are on IB Lite then you'll get filled at a better price than the best offer, not pay any commissions/exchange fees, and not cause any immediate market impact. Since the firms are assuming that the trades are zero alpha they aren't going to immediately go into the market to exit the position since they would then have to buy at a higher price, pay the 30 mil exchange fee for taking liquidity on top of what they paid for the flow.

The above does assume there isn't any midpoint/hidden liquidity within the spread available in which case accessing that liquidity is better than going to HFT.

I don't think paying for order flow is achieving best execution for clients but I do think there is a chance for a subset of traders to benefit from the arrangement.

 

[tommcginnis]

If a market is 20.00 BID and 20.25 ASK, and I go in with a 20.05 order to buy, I'm lessening the hit to those who might sell @ market by 5¢.

If some CLMF (co-located MFer) senses a movement to 20.00 to 20.05 and steps in front of my sent-yet-not-quite-arrived BID and takes the only available inventory, then yeah, I might have to go to 20.10 from 20.05, taking a shiny nickel out of my trade.

And had I wished to immediately sell -- maybe a nickel inside of 20.25, at 20.20, suddenly my prior harvest of 15¢ off the 25¢ BID-ASK spread, is now reduced by 33% to the dime between 20.10 and 20.20. Well! That sucks!

And yet, that is ALSO what happens in competitive environments -- people stepping in front of others. So if I wished to buy or sell at the MID, and the first lot goes to some CLMF who stepped in front -- what about the parties selling?? What if a CLMF stepped in front of the party offering to SELL at MID, and got me (as second contract up) in/out a nickel cheaper?? And suddenly, I'm a nickel richer (or left altogether indifferent).

The ability to co-locate and gain an undoubted time advantage does not remove the risk of being, in fact, too early -- and having someone else step in front on the other side, and move things down anyway. Anyone who's trades go longer than 2 seconds (or, 2,000 milliseconds) really doesn't have a nickel to worry about.

 

[ETJ]

What does preferenced liquidity do to the argument against PFOF. No dark pools in options, but dark liquidity as preferenced and upstairs trading is huge and in some products rivals the listed trading. You want to go where the liquidity is and price is only a portion of the equation. If I get a great fill and the market moves $.10 against me with no change in the underlying it was a shit fill. PFOF and HFT can be mutually exclusive, but they do often overlap.
16 exchanges and about half the MM is three firms and most of the other liquidity - not all - is me too quoting.

The SEC should kill PFOF, preferenced liquidity and monopoly products and private feeds that are faster than the SIP, but it won't although they have been studying maker/taker and could possibly alter that at some future point.

Discussions of PFOF and HFT trading should be had in the context of understanding preferenced liquidity. If PFOF gives me access to preferenced liquidity than it may be worth it.

 

[qlai]

I was speaking to someone senior at one of the HFT firms and they were complaining that the retail designation is too broad. They want to segregate infrequent from frequent retail like is done with options. In the options market if you place more than 390 orders on average per day then you're classified as a professional even if trading in a personal account and not affiliated with any financial firm.

Really? I would assume that any order coming from non-collocated customer would be welcomed. I would not be surprised if brokers sending orders to HFTs intentionally use the slowest equipment ( and/or HFTs intentionally use slowest equipment from/to brokers). This is what bothers me about PFOF(and internalization for that matter) - no accountability as far as latencies are concerned. Without this benchmark, no improvements can be expected, imho. The way it stands now, it's in everyone's interest, except for retail traders of course, to delay orders as much as possible. Crazy.

 

[dinn13]

Really? I would assume that any order coming from non-collocated customer would be welcomed. I would not be surprised if brokers sending orders to HFTs intentionally use the slowest equipment ( and/or HFTs intentionally use slowest equipment from/to brokers). This is what bothers me about PFOF(and internalization for that matter) - no accountability as far as latencies are concerned. Without this benchmark, no improvements can be expected, imho. The way it stands now, it's in everyone's interest, except for retail traders of course, to delay orders as much as possible. Crazy.

Colo doesn't matter per se. There are plenty of firms and specifically broker algos that are colocated that don't need to be since they aren't really making use of the lower latency. If you don't have short term alpha then latency doesn't matter. If the source of the order isn't collocated but there is a lot of size behind it then HFT or anyone that knows what they're doing doesn't want to be on the other side of it especially initially. Would want to take the other side once the order is over and mean reversion from the market impact kicks in though.

The best example would be a large vwap as a percent of adv from an asset manager. The child orders from that are going to have bad markouts for someone taking the other side (regardless of latency) so if an HFT detects it then they'll stop adding liquidity to it. HFT firms aren't obligated to fill broker algos when pinged over their SLP so as soon as they see bad markouts they'll stop adding liquidity to the client. I don't know all the details about PFOF but I assume they are obligated to fill all flow they are getting from the etrades of the world and so because of that have to eat losses from anyone with bad markouts which could be a good day trader.

I think you're making a good point as far as transparency around PFOF. Ideally there should be stats around latencies (especially how long it took for the firm to fill the order after accepting it). I have no idea what kind of information is exposed to clients but ideally there would be TCA analysis around it to attempt to verify best execution.

 

[blueraincap]

where can i find a gentle introduction to up-to-date equity market structure, including HFT implications? Larry Harris is out-of-date

 

[qlai]

If the source of the order isn't collocated but there is a lot of size behind it then HFT or anyone that knows what they're doing doesn't want to be on the other side of it especially initially. Would want to take the other side once the order is over and mean reversion from the market impact kicks in though.

Why would they need to take the other side of such flow when they can simply latency arb it against SIP? Since, as you mentioned, HFTs can identify which customer's flow is toxic then should not be a problem to them - worst case they can simply pass through. So I guess my question is - why is it not enough to know that order is allowed to be delayed up to N milliseconds to make money on it (or at least not loose any)?
Edit: is it because they are paying for it, so need to at least cover costs?