Trading with automation (with IB) - II - spreads

[fullautotrading]

Just as with Photoshop, the possibility of "logical layering" gives a great number of practical advantages. Two of the most obvious are, for instance:

- Possibility to define "hedgers", whose action can be followed/tuned separately (we have already seen that)

- Significant reduction of margin requirements

One practical example of the latter is shown, for instance, by the current 3 layers of SIlver:

If we were trading the instrument on 3 different accounts, we would have the combined margin requirements of +9, -15, -2 (total 26) contracts respectively.

Instead, by letting the software "channel" all the action on the same account - while it takes on all the work of accounting the trades on separate "logical" layers - we have a short exposure of only 8 contracts on the actual account:

But, as far as the final PNL results are concerned, they would be identical if we were actually acting long/short on 3 different accounts.

Another useful conceptual device is the "player splitter", which allows to redistribute the components of losses to be recovered into smaller parts, more suitable to smooth the hedging action.

 

[fullautotrading]

Today we are taking some "beating", with CL spiking up:

So far nothing really worrying, as we are actually "loading up" for new profits (see G-L curve, green).

On the other hand, we hold a good position on CL PUTs (e.g.: CL FOP 20180417 65 P NYMEX 1000 Light Sweet Crude Oil [ LOK8 P6500 ] ) and they are being pushed to near zero value (0.24, 0.25 right now for the PUT 65 above), so at least we have always reasons (or layers) to "comfort" us.

 

[fullautotrading]

We are back in good shape ($1.3M), having recovered most of the large DD (or "load up") and currently hammering CL the way down:

The G-L curve is particularly high (denoting a good "potential") while our "FullMaintMarginReq" is $1,863,898.22.

Current situation instrument by instrument:

I am also still continuously rolling over the CL calls some of which are in the money (at most 4%), and they are bringing in good (extra) cash due to decay.

The PUT 65 particularly, mentioned in the previous post, has made over $130K (this was rolled over as follows:
CL FOP 20180315 64 P NYMEX 1000 → CL FOP 20180315 65 P NYMEX 1000
CL FOP 20180315 65 P NYMEX 1000 → CL FOP 20180417 65 P NYMEX 1000
CL FOP 20180417 65 P NYMEX 1000 → CL FOP 20180615 65 P NYMEX 1000

note that we are free to "transfer" the trading info even across different strikes: no problem.)

 

[truetype]

When are you moving from papertrading to actual?

 

[fullautotrading]

When are you moving from papertrading to actual?

We do real trading (with private investors sufficiently capitalized).

In our illustrative threads (this and the previous ones, including many unsuccessful), however we use a paper trading account, as usually investors wish to protect their identity and what they make. [On the other hand, obviously results do not depend on the nature of the money "attached" to an account ("paper" or "real"), because the algorithmic platform is not using that information.]

None in his right mind making (significant) money in the mkt is going to show their asset on public forum. If you mean that I should disclose what I, personally, make, I do not think it is going to ever happen. Or do you want me to forward directly this info, along my SSN, to the tax officers ?

In any case, what I make or do not make is largely irrelevant. As a researcher and also developer, I am showing an approach to trading based on several years' work, based on a non-predictive approach which leverages on the the trading info (totally original as far as know) and is not relying on "guesses" or illusions of "predictions", signals, etc.. It's up to anyone who understands what I presenting and, especially, has enough capital to trade with this methodology, to find his own way.

Obviously, in case it were not clear, this is not a method to become rich, but it is a methodology to manage (existing) wealth. I am repeating this because I am constantly bombarded by people who think their will "get rich" even starting under capitalized. But sorry no. At least I personally do not have, at the moment, a legal, systematic, algorithmic method to help in that regard (and it's likely it does not even exist).

I understand this can hurt the expectations and hopes of many (perhaps including you?), but this is all I got so far.

 

[truetype]

None in his right mind making (significant) money in the mkt is going to show their asset on public forum. If you mean that I should disclose what I, personally, make, I do not think it is going to ever happen. Or do you want me to forward directly this info, along my SSN, to the tax officers ?

Yeesh. I was just asking if you'd begun real life trading or not.

 

[fullautotrading]

Yeesh. I was just asking if you'd begun real life trading or not.

I like to answer each question. Even when made for trolling/discredit purposes, because there are reading people genuinely interested to some aspects of the question.

I hope I have cleared your doubts in regard. The short answer to your question, in case it is still hard to infer from my previous reply, is yes, I would. In fact, I have already begun "real life" trading (as you call it).

If, instead, you mean whether I would begun (in the future) showing here, on a public forum, a real $ acct and the $ being made, the answer is (obviously) generally no, ... well, unless, you voluntarily provide your own account (sufficiently capitalized) for a public thread illustration, in which case we might make an exception

 

[fullautotrading]

We have gone through a new cycle of DD (or "load up", as it is preferable to see it, in this approach), even sharper than the previous one, but still not worrying.

Most of the damage to the smoothness of the PNL curve is still made by the "marginal" strategy (in "semiautomated" mode) we are playing on options. As we have already noted previously, it is a source of much greater variance for the PNL, but, the good news is that, on the other hand, the DD tend to be short lived, as the time decay is quite implacable.

This was the situation instrument by instrument at a DD point:

As it can be seen, most of the "beating" was taken obviously by options (CL calls). This is due to the fact that the semiautomated approach (based on the so called "enqueued orders") we are now using does not use hedging orders and therefore each "unfavorable" move is "suffered" in full (clearly mitigated by the long players triggered on the futures). Here is an example of outcome on the (so far) most productive option (CL put 65):

On the other hand, the fully automated algorithmic trading on futures, is proceeding quite fine. With a good long/short scalping/hedging action. (I am looking forward to remove NQ from the folio, which does not seem to add anything good to what we already have through ES.)

 

[fullautotrading]

Today we have hit a new high-water mark ($1.5M), after two powerful swings down of the PNL (emphasized by our options):

Currently, PNL is about $1.4M after about 107 solar days, 3,211 filled orders, comms: $31,169.77K.

Instrument by instrument the situation is as follows:

I am looking forward to getting rid of NQ, and, to this purpose, I have set some triggers to take profit on a favorable PNL "spike" and then shut all layers down.
I just do not like the tickdata microstructure, not to mention the options with ridiculous spread:

[ this is, in particular: NQ FOP 20180430 7000 C GLOBEX 20 E-mini NASDAQ 100 Futures QNEJ8 C7000, about -7.6% OTM ]

 

[fullautotrading]

We have been running our logic for about 16 weeks now, on a "suitable" set of instruments. It seems that the best folio would include liquid stuff such as ES, CL, SI ... and maybe a few others.

It turns out, that, instead of introducing other instruments, not really suitable for algorithmic trading (such as instruments with large spread, or with obvious drifts/decay or generally unsuitable price curves for various reasons, such as most STKs, FX etc.), it is better to use multiple layers of those which works best.

As explained before, this also allows to reduce significantly the margin requirements, due to the simultaneous long/short positions taken by the distinct layers, and maintained, at a logic level, by the platform.

Anyway, we ended up the week in good shape, with PNL fluctuating around $1.3M. The PNL curve has been fluctuating a bit more than a purely algorithmic approach (which would be generally unbeatable in this regards, as it is designed to do mainly that), by the presence of some position on options. Anyway, when there is enough capital available, it seems fine to venture in that territory too, "playing around" the algorithmic games, and grab the time decay. After all we have been using, as Full Maint Margin Req, at most $2,855,712.89 while currently we are using less than $2M (precisely: $1,930,678.05). So every "risk" has been taken so far within a quite large margin of safety.

During the weekend I am running, as usual, long simulations to improve further the current "trading games" and refining implementation details (rollover, catch up with API changes, etc.). If I find a significant improvement, I may be switching to a slightly different scalping/hedging mechanism.

As selection criterion, I usually look at the whole distribution [over multiple simulations spanning over many decades of tickdata, generated by mixtures of the most disparate stochastic processes and volatility scenarios] of my preferred performance indices (which as mentioned before, do not include the Sharpe ratio).