- ago
Sometimes there are really impressive results in some of my strategies. The "best" was an APR with 7.790.495.539.62 % (really, I made a screenshot)
This is obviously BS, luckily an easy to detect one.
A pattern could be, the sound of the finished BT is missing. It does no sound at all, at this moment I restart WL.
Mistakes from false datasets or data seem not to be the problem.
It does not enlarge my trust ;-)

0
409
38 Answers

Reply

Bookmark

Sort
- ago
#1
There is not enough details to tell what you did wrong. As always, please provide some insight into the issue: Strategy or sample code used + Strategy settings. Screenshot didn't attach, you might want to retry.
1
Cone7
- ago
#2
Before jumping to unfounded conclusions, let us give you an example that you can believe.

Consider this simulated trade I did today:



If you had put 100% of your equity into that trade, AND, the simulation was for 1 day, what would be the APR of that backtest?

Correct Answer: APR 2.2638624768198562E+23

In other words, if you could increase your equity by 17.5% every day for year, that's what the APR actually is.
1
- ago
#3
Throwing out vague accusations without providing any concrete documentation or anything to reproduce the supposed issue is kind of annoying. As Cone said, such an APR is certainly possibly, but not statistically significant.

The reason your strategy does not make a sound when it finished is that it did not generate signals. It’s the signals that produce the sound.
0
- ago
#4
My strategy converted from another platform is also getting extremely large profit and APR, at least an order of magnitude higher than on the other platform. This is causing me to perform many hours of analysis and delaying making any trades.

Any further explanation of this phenomenon would be greatly appreciated.
0
- ago
#5
Maybe you did something wrong in the strategy conversion?
0
- ago
#6
Yes, that is certainly possible.
0
- ago
#7
Rather than try for to convert a complex strategy at once you could try in stages. After each major concept is converted do a sanity check to make sure things look like they’re in line. If you have any specific code that you feel is producing incorrect results you could share it. Short of that we can only speculate 🤷🏼‍♂️
1
- ago
#8
It could be the position bug that they're fixing in Build 8 as well. If you have positions of the same size on the same symbol that close on the same bar the positions get mixed up.
0
- ago
#9
Eugene, could I send the strategy to you by mail? what email adress could I use?
0
- ago
#10
Its not a single trade, it is a minimum of 733568318,2 trades.... this is only the most ridiculous result.... not only in one strategy, but with several this problem occured.
Not only I got this problem, I know another guy in germany with sort of same problems... (ad it occured in WL 6.9. , too, in both users)


0
- ago
#11
Looks like a Look-Ahead Bias)
0
- ago
#12
QUOTE:
Looks like a Look-Ahead Bias)

Agreed. As can be seen, the Benchmark column has a mild APR which looks realistic. Which suggests a strategy programming error.
0
Cone7
- ago
#13
If you have any parameters that look like idx + n, or if you're using the << operator and then access that series in the trading rules, you're peeking. Simple as that.

There are dozens of ways to shoot yourself in the foot. If you can't figure it out, show us any strategy that does this, and we'll quickly point out the error.
0
- ago
#14
I figured out that my big difference is apparently due to NSF's and position sizing.
0
- ago
#15
No, this example is not a new strategy with always strange results, it is the sample knife juggler just with a transaction weight and a stop loss exits added. And with different data set.
The problem occurde several times with several strategies.
Most times there are realistic results, but sometimes with these strange results.
At least sometimes with signals but without sound.
After restart wl the "normal" results (with bt-signal-sound) are back...
Some problem with european computer (as a german collegue had simular probs)?
the rest in next picture

0
- ago
#16
0
- ago
#17
QUOTE:
Its not a single trade, it is a minimum of 733568318,2 trades

What do you mean, are there fractional "733568318,2" trades?

For the record, what regional settings are selected in Windows?
0
- ago
#18
Maybe we can schedule a screen share so you can demonstrate the issue in real time?
0
- ago
#20
QUOTE:
What do you mean, are there fractional "733568318,2" trades?

It was the excel result APR / Profit Per Bar. No fractional trades, sorry for the irritation.

QUOTE:
For the record, what regional settings are selected in Windows?

region: Gemany
Regional format: recommended: Germany
time zone UTC +1 (Berlin, Amsterdam)
0
- ago
#21
maybe here?
0
- ago
#22
Thanks, you're using German settings with "," being the separator char. We tested this regional setup extensively so it should be fine. Maybe a screen share as Glitch suggests should help troubleshooting.

P.S. As JPG is truly suboptimal for screenshots, you might want to give PNG a try.
0
- ago
#23
QUOTE:
screenshare

of course, no problem. When? how?

QUOTE:
P.S. As JPG is truly suboptimal for screenshots, you might want to give PNG a try.
OK, I found a way to safe it in PNG
0
- ago
#24
Screenshare on monday morning us time?
Zoom? Anydesk? Skype?
Best to contact me info at sts-am dot com
0
Cone7
- ago
#25
I'll send you a Zoom invite.

What's the DataSet and Provider you're using with 1595 symbols?
I'm trying to duplicate this now.
0
- ago
#26
Data set are the biggest stock companies in USA, selected at zacks.com. I can send you the symbols by email....
0
- ago
#27
You could attach it as a text file here?
0
- ago
#28
A AAGIY AAL AAPL ABB ABBV ABC ABEV ABMD ABNB ABT ACGBY ACGL ACN ACWI ADBE ADDYY ADI ADM ADP ADRNY ADSK AEE AEM AEP AES AFL AFRM AGG AGR AHKSY AIG AIQUY AJG AKAM AKZOY ALB ALC ALGN ALIOF ALIZY ALL ALLY ALNY ALPMY ALSMY ALXN AMADY AMAT AMCR AMD AME AMGN AMKBY AMP AMT AMX AMZN ANET ANPDF ANSS ANTM ANZBY AON AONNY APD APH APTV ARE ARES ARGX ASAZY ASHTY ASMIY ASML ASX ATASY ATLKY ATUS ATVI AVB AVGO AVIFY AVTR AVVIY AVY AWK AXAHY AXP AZN AZO BA BABA BAC BACHY BAESY BAM BAMXF BASFY BAX BBD BBDO BBVA BBY BCE BCMXY BCS BDORY BDX BEKE BEN BF.A BF.B BGNE BHKLY BIDU BIIB BILI BIO BIO.B BIP BK BKNG BKR BLK BLL BMO BMRN BMXMF BMY BNPQY BNS BNTGY BNTX BOIVF BOUYF BP BR BRDCY BRK.A BRK.B BRO BSBR BSX BTI BUD BURL BX BXP BYDDY C CABGY CAG CAH CAJ CARR CAT CB CBRE CCEP CCI CCK CCL CDNS CDW CE CEO CERN CFG CG CGEMY CGNX CHD CHDRY CHKP CHTR CHWY CI CIADY CICHY CIHKY CINF CJEWY CJPRY CKHUY CL CLPHY CLVT CLX CM CMAKY CMCSA CME CMG CMI CMPGY CMS CMWAY CNC CNHI CNI CNQ COF COO COST COUP CP CPB CPNG CPRT CQP CRARY CRH CRHKY CRL CRM CRRFY CRWD CS CSCO CSGP CSLLY CSUAY CSX CTAS CTLT CTPCY CTRYY CTSH CTTAY CTVA CTXS CVAC CVE CVNA CVS CVX CZR D DAL DANOY DASH DASTY DB DBOEY DBSDY DCMYY DD DDAIF DDOG DE DELL DEN DEO DFS DG DGX DHI DHR DIA DIS DISCA DISCB DISCK DISH DKILY DKNG DLR DLTR DNHBY DNKEY DNZOY DOCU DOV DOW DPSGY DPZ DRE DRI DSDVY DT DTE DTEGY DUK DVN DVY DWAHY DWHHF DXCM EA EADSY EBAY EBKDY EC ECIFY ECL ED EDPFY EDU EEM EFA EFX EIX EL ELAN EMB EMN EMR ENB ENGIY ENLAY ENPH ENTG EOG EONGY EPAM EPD EQH EQIX EQNR EQR ERFSF ERIC ES ESALY ESLOY ESS ET ETN ETR ETSY EVRG EW EXAS EXC EXPD EXPE EXPGY EXR F FANUY FAST FB FBHS FCX FDX FE FERG FICO FIS FISV FITB FJTSY FLT FMC FMS FNV FOX FOXA FQVLF FRC FRRVY FSNUY FSUGY FTCH FTNT FTS FTV FUJHY FUJIY FUTU GBOOY GD GDDY GDRX GDS GE GELYY GH GIB GILD GIS GLD GLNCY GLW GM GMAB GNMSF GNRC GOLD GOOG GOOGL GPC GPN GRMN GS GSK GVDNY GWW GZPFY HAL HBAN HCA HCMLY HD HDB HDELY HEI HEI.A HEINY HENKY HES HIG HLDCY HLT HMC HNNMY HOCPY HOKCY HOLX HON HPE HPQ HRL HSBC HSY HTHT HUBS HUM HVRRY HWM HYG HYMLF HZNP IAC IAU IBDRY IBKR IBM IBN ICE IDCBY IDXX IEF IEX IFF IFNNY IGSB IJH IJR ILMN IMBBY IMO IMPUY INCY INFO INFY ING INTC INTU INVH IP IQV IR ISNPY ISRG IT ITOCY ITUB ITW IVV IWD IWF IWM IWN IWP IWR IX J JBHT JBSAY JCI JD JHX JMHLY JNJ JPM JSHLY K KB KBCSY KDDIY KDP KEY KEYS KHC KKR KLAC KMB KMI KMTUY KMX KNBWY KNYJY KO KR KRYAY KSU KUBTY KYCCF KYOCY L LB LBRDA LBRDK LBTYA LBTYB LBTYK LDOS LEN LEN.B LFC LGGNY LH LHX LI LIN LLY LMT LNG LNNGY LNT LNVGY LOGI LOW LQD LRCX LRLCY LSXMA LSXMB LU LUKOY LULU LUV LVMUY LVS LYB LYFT LYG LYV LZAGY MA MAA MAR MARUY MAS MBB MCD MCHP MCK MCO MDB MDLZ MDT MELI MET MFC MFG MGA MGDDY MGM MHK MIELF MITEY MITSY MKC MKL MKTX MLM MMC MMM MNST MO MOH MPC MPLX MPWR MRAAY MRK MRNA MRVL MS MSADY MSBHF MSCI MSFT MSI MT MTB MTCH MTD MU MUB MUFG MURGY MXIM NCMGY NDAQ NEE NEM NET NFLX NGG NGLOY NICE NILSY NIO NJDCY NKE NMR NNGRY NOC NOK NOW NPSCY NPSNY NRDBY NRILY NSANY NSC NSRGY NTAP NTDOY NTES NTIOF NTR NTRS NTTYY NUAN NUE NVAX NVCR NVDA NVO NVR NVS NVZMY NWG NWS NWSA NXGPY NXPI O OAS OCDDY OCPNY ODFL OGZPY OKE OKTA OLCLY OMC OMRNY ON ORAN ORCL ORLY OSH OTIS OVCHY OXY PANW PAYC PAYX PBA PBR PBR.A PCAR PCG PCRFY PDD PDRDY PDYPY PEAK PEG PENN PEP PFE PFF PFG PG PGR PH PHG PHM PINS PKI PKX PLD PLTR PM PNC PNGAY POAHY PODD POOL PPD PPERY PPG PPL PPRUY PROSY PRU PSA PSMMY PSX PTC PTON PTR PUBGY PUK PXD PYPL QCOM QQQ QRVO QSR RACE RANJY RBGLY RBLX RCI RCL RCRRF RDS.A RDS.B RDSMY REGN RELX REPYY RF RHHBY RIO RJF RKT RKUNY RLX RMD RNG ROK ROKU ROL ROP ROST RPRX RSG RSP RTX RWEOY RY RYAAY SAFRY SAM SAN SAP SBAC SBGSY SBMFF SBRCY SBSW SBUX SCBFF SCCO SCGLY SCHW SCHYY SCMWY SDVKY SDXAY SDY SE SEGXF SEOAY SFTBY SGEN SGSOY SGTZY SHECY SHG SHOP SHV SHW SHY SIEGY SIRI SIVB SJM SKHSY SKM SLB SLF SMCAY SMFG SMPNY SNAP SNN SNOW SNP SNPS SNY SO SOMLY SONVY SONY SPG SPGI SPLK SPOT SPY SQ SRE SSEZY SSMXY SSNC SSREY SSUMY STE STLA STM STNE STT STX STZ SU SUHJY SUI SUZ SWDBY SWGAY SWK SWKS SWRAY SYF SYIEY SYK SYY SZKMY T TAK TCEHY TCOM TD TDG TDOC TDY TEAM TECH TEF TEL TER TEZNY TFC TFX TGT TIP TJX TKOMY TLK TLSNY TLSYY TLT TM TME TMO TMUS TNABY TOELY TOSYY TOT TRI TRMB TROW TRP TRU TRUMY TRV TSCDY TSCO TSLA TSM TSN TT TTD TTDKY TTNDY TTWO TU TW TWLO TWTR TXG TXN TYL U UAL UBER UBS UI UL ULTA UMC UNCFF UNH UNICY UNP UPS URI USB V VALE VCISY VEEV VEOEY VFC VIAC VIACA VICI VIPS VIVHY VLO VLVLY VMC VMW VNQ VOD VONOY VRSK VRSN VRTX VTI VTR VTRS VWAGY VWDRY VWO VZ W WAB WAT WBA WCN WDAY WDC WEC WEICY WELL WFC WHR WIT WIX WLMIY WLTW WM WMB WMG WMT WOPEY WORK WPM WPP WRB WRK WST WTKWY WY WYNN XEL XLF XLNX XLY XM XNGSY XOM XP XPEV XPO XRAY XYL YAHOY YARIY YNDX YUM YUMC Z ZBH ZBRA ZEN ZG ZI ZLAB ZLNDY ZM ZS ZTO ZTS ZURVY
0
Cone7
- ago
#30
Which provider are you using for the DataSet?
If Unlinked, which Provider do you have checked at the top of the list?
0
- ago
#31
afaik unlinked, wealthdata, q-data, alphavantage, yahoo, CBOE, Nasdaq, IB,

with data extention by rene koch

today there was an error with yahoo data.. now it seem to work again without error...
but 2 strategies have different values, 2 strategies have the same value as this morning... but VERY diffent to the values on friday and early morning (you can see screenshots in our online session)
0
Cone7
- ago
#32
Using the symbol list, I already see the problem. Wealth-Lab is allowing the backtest to buy unrealistic amounts, volume-wise, of penny stocks.

What you should do is go to Preferences > Backtest > Other Settings and enter a non-zero number for Volume % Limit. This won't block trades, but will limit the volume you can purchase - to add some realism to volume of the purchases.

Give that a try, and let us know how it works out for your backtest.
0
- ago
#33
well, thanks, that is a step in the right direction.
it shows the same results in both tests, but surprisingly with an irritating number of positions.
Solution?? I deleted all local data from all dataprovider - now it seems to be more realistic in this test. I will do some more tests with other parameter

0
- ago
#34
(fyi: build 8, all extentions updated, all local data from providers deleted a few days ago)

Same strategy sts knife juggler with transaction weight like obove, but different dataset:
Backtest 6 month:

Imported stocks from zacks.com APR 425 %
US Russell 1000 (1017 symbols) by yahoo (should be mostly same companies) APR 29,74
US Russell 2000 by Yahoo APR 29,74 %
SP500 by yahoo APR 71,94 % (this seems to be strange to me...)

Datasets seem to be a great effector, but whats wrong with imported data symbols? Cant I trust selfbuild datasets? Did I make a mistake by creating my own dataset?? Is there a mistake in the selection of data providers?


0
- ago
#35
It's commonplace but backtests on different data are likely to be different. Bad bars, curated bars, excessive splits, missing splits, the list can go on and on. If I were in you position I'd start by reviewing each trade with a suspicious PnL.
0
- ago
#36
Sorry Eugene, please read my 2 postings and data again.
Making just 4 trades in 4 years and anopther run with almost 4000 trades has nothing to do with dataset or splits (how many split did occur in 6 Month?)
APR between 30% and 425% in 6 month with almost same data (Russell 1000 should not be too far away from 1007 symbols the biggest Companies in USA)


0
- ago
#37
I can only speculate how you could do own troubleshooting with the free data you've downloaded from various data vendors 🤷‍♂️
0
Cone7
- ago
#38
This is hard to follow. Only 4 trades but you're using S&P 500 and Russell x 1000 indices?
What does that mean?
0

Reply

Bookmark

Sort