Question for Grob/Hershey...

Status
Not open for further replies.
Quote from txuk:

For the last fifteen minutes I have been running the IQFeed and IB spreadsheets side-by-side. The vast majority of the time they are step-in-step, within a split second. But I did notice occasional lag on the IB, which could cause for missing some of the extremes. Example, both show a squ of -3, then IQ quickly jumps to -3.4 and returns to -3, while IB sticks on -3.
More...

Some technicalities as my YM is not updating... If you look at rows 54 columns G and H, you'll see two timestamps for you IQFEED DDE folks. I had notice from time to time that IQFEED can occasionally hang. When I see prolonged periods of residing in STR (ie. GREEN ZONE) or SQU (ie. RED ZONE), I check to make sure that the datafeed is updating. Because these are liquid pairs, the time stamped in those two cells should be current. If one of them aren't, then I close the sheet and reload it.

If you recall from a late post last night, STR/SQU do not prolong themselves for long periods of the time. The cycling of NEUT/STR and NEUT/SQU do but not the SINGLE STR or SQU states. The reason why is because of the arb folks. They are a swift bunch of folks who quickly seize the opportunity for as long as it persists. They are also largely responsible for driving out the ability to backtest STR.SQU in addition to two other technical factors. If you see a STR or SQU in excess of 90 seconds, it is a good idea to consider recentering as described a few posts back. I am loose in addition to data gathering my way down to this level of fineness so I am rarely bothered by it. The recentering is seldom in excess of a +/- 1 adjustment to the market offset we plug in daily.

For txuk, there is a very technical difference between IQFEED and IB. If you watch ER2 it is noticeable very often especially given the NOW screen I use. IB is snapshot driven. The rep rate is something like every 300 ms. Occasionally IQFEED will hang especially when PRV redlined (just had a thought, a PRV tachometer would be a super visual tool). The advantage of IB's snapshot is that when things are really moving, there is no performance drag in IB's ability to update information. With IQFEED, it will lag when the asset is really moving due to it's pipeline framework.

The other technical point about the sheet is that the realtime NEUTRALIZED offset is calculated by taking the difference between the last YM price and INDU price less the MARKET OFFSET (ie. the number we plug in every day). Like all assets, the last price value actually belongs to a pair that is either the BID PRICE or the ASK PRICE. For us the difference between the pair is 1, the Bid/Ask YM spread. Again these are just very fine details which do not bother me. Perhaps using the average between the bid/ask spread would suit you better as opposed to using the last price...

The syntax would be as follows

replace "=IQLink|'@YMM6'!last"

with "=(IQLink|'@YMM6'!bid+IQLink|'@YMM6'!ask)/2"

Again, I am a loose person and the original sheet is what I am calibrated with. However, the sheet was built to cater to the flexibility of most... Change the colors, do what you please.

Kind Regards,
MAK
 
Ok so for the IB xls I posted last night, the timestamps Mak refers to are removed - they were somehow causing me problems in mine. That is a clever way Mak to keep tabs on how up to date the data feed is.

Not sure if IB has a timestamp function - if anyone has it working please advise

IRL


Quote from makosgu:

If you look at rows 54 columns G and H, you'll see two timestamps for you IQFEED DDE folks. I had notice from time to time that IQFEED can occasionally hang.
More...
 
Just an fyi... I am running a 57 for the smart money offset as opposed to the arbitrageurs 59 (indexarb.com) since I didn't seam to mention this in an earlier post... Recall that we can begin with the indexarb value or even the previous days offset. Upon NEUTRALIZING (ie. SYNCH), we then get calibrated to the SMART MONEY on the YM futures. However, don't trust me, trust what you see yourselves and debrief any probs...

Grob, if you're trading today, it would be interesting to compare the offset you're pinning down. Perhaps I can incorporate the variations that txuk refers to.

I am not continuallyglued to the STR.SQU like some of you may be are. That is what sweeping is for... Today has thus far been nicely catering to PAIR TRANSLATIONS and STR.SQU analysis of 2 PAIR STALLS/RESUMES/REVERSEs. Sweeping a PAIR TRANSLATION is a no change. This is a later down the line type of stuff... Getting calibrated to the DATA GATHERING and ANALYSIS of STR.NEUT.SQU is the job at hand.

Regards,
MAK
 
Circa 3:40 EST the FV did move out to 59. I was immediately tipped off by the shift when a short translation began on a STRetched/2 Pair. YES, I am not PERFECT... :cool:. On top of that, EOD activity begins to influence the market. For some brokers, intraday margins end at 3:45 etc... By 4:00, INDU.X is flatlined whilst futures action continues. We get all the EOD effects pushing in to close the market.

A dozen or so have PM'd me about a whole myriad of things. It is best to post things in the forum. You'd be surpise how many duplicate responses I am often PMing right back out. As usual, I don't regard any of this as an edge so I just lay out the whole nine yards that I currently see. Q's about how to use the sheet have been hashed repeatedly for the last 3 days. Most of the PM'd Q's are already answered in the thread. I have taken a CONSIDERABLE amount of time (months) to strip things down to the bare bones that only requires the user to do two things, 1) enter an initial offset 2) calibrate the initial offset to the smart money offset once synch has occurred and occasionally recenter as a third item.

Believe it or not, I still keep what most of my close circle of friends call a very high profile day job as a "Vice President" of a demanding start up. My daily take home salary from the job is considerably more than what the FOUL posted yesterday. It is both a blessing (trade-learning wise since it allows trading to be pressureless) and a curse (time wise). It is just another one of those avenues in which I make a strong effort to contributing to getting a job done (hopefully, yet another IPO :cool: ). My trading experience to date is now just a bit over a year; hence my consideration of myself as a peer amonst peers. And so tommorrow is yet another day which will continue to reinforce that which some of you are beginning to see and get tuned in to.

Kind Regards,
MAK
 
Mak, as you've said, the indexarb FV is fairly accurate out of the box, but I've been thinking about our need to occasionally make manual adjustments during the day based on observation. Have you considered calculating FV dynamically?

"The fair value premium equals the interest earned on the spot index minus dividends earned by the stocks that comprise the index. The relevant time period is from the current date until the future's expiration."

Their formula is: FV = Index Value * [(1 + interest rate)(Number Days / 365)-1] - (Sum of Dividends) / Divisor

I'll assume the dividend and divisor are fairly constant intraday, but interest rate and time remaining are constantly changing. So it makes sense that the early AM and late PM should have different FVs, especially if there was a significant swing in interest rates during the day. Indexarb's formula has a precision of one day, but could be easily adjusted down to minute. It seems the only hurdle to calcing dynamically is getting an accurate up to the minute interst rate.

They explain: "The interest, or cost of carry, component is based on interest rates determined from a zero coupon yield curve that is constructed from Deposit rates and Eurodollar futures. The interest rate that corresponds to the exact number of days remaining in the futures contract is determined from the yield curve using interpolation. This method is used to value swap instruments and is a widely used method for determining rates to discount cash flows."

This has probably already been discussed on ET, I didn't do a search. But I'm curious if you have explored this yet?
 
Quote from txuk:

Mak, as you've said, the indexarb FV is fairly accurate out of the box, but I've been thinking about our need to occasionally make manual adjustments during the day based on observation. Have you considered calculating FV dynamically?
More...

In the PM, I only noted a single update was warranted as per EOD activity firing up. I know alot of others experience a lack of confidence when they start dealing with what ifs... The tiniest of things will begin to bother you and impair the fact that your actual action was correct.

I have toyed with the solution you propose but opted otherwise for various reasons. The premier fact being that SMART MONEY will establish the true offset as opposed to the theoretical FAIR VALUE. If you google around, you will find several other sites that post the theorhetical FV as they see it. Squawk Box on MSNBC reports the FV everyday before the markets open. Often, all 3 disagree. Look at indexarb, they report the noise range as being between +/-10... ??? As a result, it quickly became evident that all 3 failed to observe and get recalibrated to the live instrument.

A while ago I contemplated coding the realtime FV which you can go ahead and give it a shot. Otherwise, there is of course a solution which requires a fair bit of code. It is complex for various reasons but not impossible I refrained from coding it because it introduces a performance lag into the sheet. DDE for whatever reason seams to maintain sufficient performance (ie. non-lag). The conditional formatting formulas have the potential to lag if they are used in excess (ie. hundreds of cells).

In any event, the coding solution is to introduce code that tracks the two pair state and flags when a pair translates counter the SMART MONEY'S STR/SQU threshold (ie. STR(+3)=>HIGHER PAIR=>STR(+2)=>LOWER PAIR=>NEUT=>LOWER PAIR). There are several other additional state combinations that would need to be flagged by the code which I'm sure you can figure out (a lengthy OR statement). Coding-wise, for me this was the only acceptable solution since it calibrates exclusively to the actual instrument as opposed to the theorhetical FV. However, because I was able to recognize at a glance when this occurs, I opted for the punch in to save performance and the headache of coding.

In a nutshell, the worksheet_calculate() pseudo-code is as follows

Sub worksheet_calculate()
IF {ANY OF THE FLAGGED CONDITIONS ARE MET} THEN
IF {FALSE STR} THEN
INCREMENT FV UNTIL NEUTRAL is ZEROED
ELSE //{FALSE SQU}
DECREMENT FV UNTIL NEUTRAL is ZEROED
END IF
END IF
End Sub

The above pseudo-code is massively simplified and understated. My recommendation would be to screen record a segment of the day and filter out the states logic. The screen recording will help to rewind and fast forward to the points that you are concerned with. There are several cheaper solutions than camtasia.

Regards,
MAK
 
Quote from makosgu:I have toyed with the solution you propose but opted otherwise for various reasons.
More...
Thanks for the detailed response, I figured you must have looked into this at some point. I'm not worried about fine-tuning the decimal points and from what I've seen using your spreadsheet I don't believe that kind of precision is necessary to determine the right side of the mkt. Jack also made that point yesterday. This was just a curiousity that struck me when I looked at the FV formula.
 
Status
Not open for further replies.
Back
Top