Historical Currency Converter — Articles — OANDA FX Data ...

H1 Backtest of ParallaxFX's BBStoch system

Disclaimer: None of this is financial advice. I have no idea what I'm doing. Please do your own research or you will certainly lose money. I'm not a statistician, data scientist, well-seasoned trader, or anything else that would qualify me to make statements such as the below with any weight behind them. Take them for the incoherent ramblings that they are.
TL;DR at the bottom for those not interested in the details.
This is a bit of a novel, sorry about that. It was mostly for getting my own thoughts organized, but if even one person reads the whole thing I will feel incredibly accomplished.

Background

For those of you not familiar, please see the various threads on this trading system here. I can't take credit for this system, all glory goes to ParallaxFX!
I wanted to see how effective this system was at H1 for a couple of reasons: 1) My current broker is TD Ameritrade - their Forex minimum is a mini lot, and I don't feel comfortable enough yet with the risk to trade mini lots on the higher timeframes(i.e. wider pip swings) that ParallaxFX's system uses, so I wanted to see if I could scale it down. 2) I'm fairly impatient, so I don't like to wait days and days with my capital tied up just to see if a trade is going to win or lose.
This does mean it requires more active attention since you are checking for setups once an hour instead of once a day or every 4-6 hours, but the upside is that you trade more often this way so you end up winning or losing faster and moving onto the next trade. Spread does eat more of the trade this way, but I'll cover this in my data below - it ends up not being a problem.
I looked at data from 6/11 to 7/3 on all pairs with a reasonable spread(pairs listed at bottom above the TL;DR). So this represents about 3-4 weeks' worth of trading. I used mark(mid) price charts. Spreadsheet link is below for anyone that's interested.

System Details

I'm pretty much using ParallaxFX's system textbook, but since there are a few options in his writeups, I'll include all the discretionary points here:

And now for the fun. Results!

As you can see, a higher target ended up with higher profit despite a much lower winrate. This is partially just how things work out with profit targets in general, but there's an additional point to consider in our case: the spread. Since we are trading on a lower timeframe, there is less overall price movement and thus the spread takes up a much larger percentage of the trade than it would if you were trading H4, Daily or Weekly charts. You can see exactly how much it accounts for each trade in my spreadsheet if you're interested. TDA does not have the best spreads, so you could probably improve these results with another broker.
EDIT: I grabbed typical spreads from other brokers, and turns out while TDA is pretty competitive on majors, their minors/crosses are awful! IG beats them by 20-40% and Oanda beats them 30-60%! Using IG spreads for calculations increased profits considerably (another 5% on top) and Oanda spreads increased profits massively (another 15%!). Definitely going to be considering another broker than TDA for this strategy. Plus that'll allow me to trade micro-lots, so I can be more granular(and thus accurate) with my position sizing and compounding.

A Note on Spread

As you can see in the data, there were scenarios where the spread was 80% of the overall size of the trade(the size of the confirmation candle that you draw your fibonacci retracements over), which would obviously cut heavily into your profits.
Removing any trades where the spread is more than 50% of the trade width improved profits slightly without removing many trades, but this is almost certainly just coincidence on a small sample size. Going below 40% and even down to 30% starts to cut out a lot of trades for the less-common pairs, but doesn't actually change overall profits at all(~1% either way).
However, digging all the way down to 25% starts to really make some movement. Profit at the -161.8% TP level jumps up to 37.94% if you filter out anything with a spread that is more than 25% of the trade width! And this even keeps the sample size fairly large at 187 total trades.
You can get your profits all the way up to 48.43% at the -161.8% TP level if you filter all the way down to only trades where spread is less than 15% of the trade width, however your sample size gets much smaller at that point(108 trades) so I'm not sure I would trust that as being accurate in the long term.
Overall based on this data, I'm going to only take trades where the spread is less than 25% of the trade width. This may bias my trades more towards the majors, which would mean a lot more correlated trades as well(more on correlation below), but I think it is a reasonable precaution regardless.

Time of Day

Time of day had an interesting effect on trades. In a totally predictable fashion, a vast majority of setups occurred during the London and New York sessions: 5am-12pm Eastern. However, there was one outlier where there were many setups on the 11PM bar - and the winrate was about the same as the big hours in the London session. No idea why this hour in particular - anyone have any insight? That's smack in the middle of the Tokyo/Sydney overlap, not at the open or close of either.
On many of the hour slices I have a feeling I'm just dealing with small number statistics here since I didn't have a lot of data when breaking it down by individual hours. But here it is anyway - for all TP levels, these three things showed up(all in Eastern time):
I don't have any reason to think these timeframes would maintain this behavior over the long term. They're almost certainly meaningless. EDIT: When you de-dup highly correlated trades, the number of trades in these timeframes really drops, so from this data there is no reason to think these timeframes would be any different than any others in terms of winrate.
That being said, these time frames work out for me pretty well because I typically sleep 12am-7am Eastern time. So I automatically avoid the 5am-6am timeframe, and I'm awake for the majority of this system's setups.

Moving stops up to breakeven

This section goes against everything I know and have ever heard about trade management. Please someone find something wrong with my data. I'd love for someone to check my formulas, but I realize that's a pretty insane time commitment to ask of a bunch of strangers.
Anyways. What I found was that for these trades moving stops up...basically at all...actually reduced the overall profitability.
One of the data points I collected while charting was where the price retraced back to after hitting a certain milestone. i.e. once the price hit the -61.8% profit level, how far back did it retrace before hitting the -100% profit level(if at all)? And same goes for the -100% profit level - how far back did it retrace before hitting the -161.8% profit level(if at all)?
Well, some complex excel formulas later and here's what the results appear to be. Emphasis on appears because I honestly don't believe it. I must have done something wrong here, but I've gone over it a hundred times and I can't find anything out of place.
Now, you might think exactly what I did when looking at these numbers: oof, the spread killed us there right? Because even when you move your SL to 0%, you still end up paying the spread, so it's not truly "breakeven". And because we are trading on a lower timeframe, the spread can be pretty hefty right?
Well even when I manually modified the data so that the spread wasn't subtracted(i.e. "Breakeven" was truly +/- 0), things don't look a whole lot better, and still way worse than the passive trade management method of leaving your stops in place and letting it run. And that isn't even a realistic scenario because to adjust out the spread you'd have to move your stoploss inside the candle edge by at least the spread amount, meaning it would almost certainly be triggered more often than in the data I collected(which was purely based on the fib levels and mark price). Regardless, here are the numbers for that scenario:
From a literal standpoint, what I see behind this behavior is that 44 of the 69 breakeven trades(65%!) ended up being profitable to -100% after retracing deeply(but not to the original SL level), which greatly helped offset the purely losing trades better than the partial profit taken at -61.8%. And 36 went all the way back to -161.8% after a deep retracement without hitting the original SL. Anyone have any insight into this? Is this a problem with just not enough data? It seems like enough trades that a pattern should emerge, but again I'm no expert.
I also briefly looked at moving stops to other lower levels (78.6%, 61.8%, 50%, 38.2%, 23.6%), but that didn't improve things any. No hard data to share as I only took a quick look - and I still might have done something wrong overall.
The data is there to infer other strategies if anyone would like to dig in deep(more explanation on the spreadsheet below). I didn't do other combinations because the formulas got pretty complicated and I had already answered all the questions I was looking to answer.

2-Candle vs Confirmation Candle Stops

Another interesting point is that the original system has the SL level(for stop entries) just at the outer edge of the 2-candle pattern that makes up the system. Out of pure laziness, I set up my stops just based on the confirmation candle. And as it turns out, that is much a much better way to go about it.
Of the 60 purely losing trades, only 9 of them(15%) would go on to be winners with stops on the 2-candle formation. Certainly not enough to justify the extra loss and/or reduced profits you are exposing yourself to in every single other trade by setting a wider SL.
Oddly, in every single scenario where the wider stop did save the trade, it ended up going all the way to the -161.8% profit level. Still, not nearly worth it.

Correlated Trades

As I've said many times now, I'm really not qualified to be doing an analysis like this. This section in particular.
Looking at shared currency among the pairs traded, 74 of the trades are correlated. Quite a large group, but it makes sense considering the sort of moves we're looking for with this system.
This means you are opening yourself up to more risk if you were to trade on every signal since you are technically trading with the same underlying sentiment on each different pair. For example, GBP/USD and AUD/USD moving together almost certainly means it's due to USD moving both pairs, rather than GBP and AUD both moving the same size and direction coincidentally at the same time. So if you were to trade both signals, you would very likely win or lose both trades - meaning you are actually risking double what you'd normally risk(unless you halve both positions which can be a good option, and is discussed in ParallaxFX's posts and in various other places that go over pair correlation. I won't go into detail about those strategies here).
Interestingly though, 17 of those apparently correlated trades ended up with different wins/losses.
Also, looking only at trades that were correlated, winrate is 83%/70%/55% (for the three TP levels).
Does this give some indication that the same signal on multiple pairs means the signal is stronger? That there's some strong underlying sentiment driving it? Or is it just a matter of too small a sample size? The winrate isn't really much higher than the overall winrates, so that makes me doubt it is statistically significant.
One more funny tidbit: EUCAD netted the lowest overall winrate: 30% to even the -61.8% TP level on 10 trades. Seems like that is just a coincidence and not enough data, but dang that's a sucky losing streak.
EDIT: WOW I spent some time removing correlated trades manually and it changed the results quite a bit. Some thoughts on this below the results. These numbers also include the other "What I will trade" filters. I added a new worksheet to my data to show what I ended up picking.
To do this, I removed correlated trades - typically by choosing those whose spread had a lower % of the trade width since that's objective and something I can see ahead of time. Obviously I'd like to only keep the winning trades, but I won't know that during the trade. This did reduce the overall sample size down to a level that I wouldn't otherwise consider to be big enough, but since the results are generally consistent with the overall dataset, I'm not going to worry about it too much.
I may also use more discretionary methods(support/resistance, quality of indecision/confirmation candles, news/sentiment for the pairs involved, etc) to filter out correlated trades in the future. But as I've said before I'm going for a pretty mechanical system.
This brought the 3 TP levels and even the breakeven strategies much closer together in overall profit. It muted the profit from the high R:R strategies and boosted the profit from the low R:R strategies. This tells me pair correlation was skewing my data quite a bit, so I'm glad I dug in a little deeper. Fortunately my original conclusion to use the -161.8 TP level with static stops is still the winner by a good bit, so it doesn't end up changing my actions.
There were a few times where MANY (6-8) correlated pairs all came up at the same time, so it'd be a crapshoot to an extent. And the data showed this - often then won/lost together, but sometimes they did not. As an arbitrary rule, the more correlations, the more trades I did end up taking(and thus risking). For example if there were 3-5 correlations, I might take the 2 "best" trades given my criteria above. 5+ setups and I might take the best 3 trades, even if the pairs are somewhat correlated.
I have no true data to back this up, but to illustrate using one example: if AUD/JPY, AUD/USD, CAD/JPY, USD/CAD all set up at the same time (as they did, along with a few other pairs on 6/19/20 9:00 AM), can you really say that those are all the same underlying movement? There are correlations between the different correlations, and trying to filter for that seems rough. Although maybe this is a known thing, I'm still pretty green to Forex - someone please enlighten me if so! I might have to look into this more statistically, but it would be pretty complex to analyze quantitatively, so for now I'm going with my gut and just taking a few of the "best" trades out of the handful.
Overall, I'm really glad I went further on this. The boosting of the B/E strategies makes me trust my calculations on those more since they aren't so far from the passive management like they were with the raw data, and that really had me wondering what I did wrong.

What I will trade

Putting all this together, I am going to attempt to trade the following(demo for a bit to make sure I have the hang of it, then for keeps):
Looking at the data for these rules, test results are:
I'll be sure to let everyone know how it goes!

Other Technical Details

Raw Data

Here's the spreadsheet for anyone that'd like it. (EDIT: Updated some of the setups from the last few days that have fully played out now. I also noticed a few typos, but nothing major that would change the overall outcomes. Regardless, I am currently reviewing every trade to ensure they are accurate.UPDATE: Finally all done. Very few corrections, no change to results.)
I have some explanatory notes below to help everyone else understand the spiraled labyrinth of a mind that put the spreadsheet together.

Insanely detailed spreadsheet notes

For you real nerds out there. Here's an explanation of what each column means:

Pairs

  1. AUD/CAD
  2. AUD/CHF
  3. AUD/JPY
  4. AUD/NZD
  5. AUD/USD
  6. CAD/CHF
  7. CAD/JPY
  8. CHF/JPY
  9. EUAUD
  10. EUCAD
  11. EUCHF
  12. EUGBP
  13. EUJPY
  14. EUNZD
  15. EUUSD
  16. GBP/AUD
  17. GBP/CAD
  18. GBP/CHF
  19. GBP/JPY
  20. GBP/NZD
  21. GBP/USD
  22. NZD/CAD
  23. NZD/CHF
  24. NZD/JPY
  25. NZD/USD
  26. USD/CAD
  27. USD/CHF
  28. USD/JPY

TL;DR

Based on the reasonable rules I discovered in this backtest:

Demo Trading Results

Since this post, I started demo trading this system assuming a 5k capital base and risking ~1% per trade. I've added the details to my spreadsheet for anyone interested. The results are pretty similar to the backtest when you consider real-life conditions/timing are a bit different. I missed some trades due to life(work, out of the house, etc), so that brought my total # of trades and thus overall profit down, but the winrate is nearly identical. I also closed a few trades early due to various reasons(not liking the price action, seeing support/resistance emerge, etc).
A quick note is that TD's paper trade system fills at the mid price for both stop and limit orders, so I had to subtract the spread from the raw trade values to get the true profit/loss amount for each trade.
I'm heading out of town next week, then after that it'll be time to take this sucker live!

Live Trading Results

I started live-trading this system on 8/10, and almost immediately had a string of losses much longer than either my backtest or demo period. Murphy's law huh? Anyways, that has me spooked so I'm doing a longer backtest before I start risking more real money. It's going to take me a little while due to the volume of trades, but I'll likely make a new post once I feel comfortable with that and start live trading again.
submitted by ForexBorex to Forex [link] [comments]

Oanda - YouTube How to Download Historial Forex Data - Metatrader 4 ... Why I now focus on Oanda forex trading API How to Build a Winning Machine Learning FOREX Strategy in ... OANDA - YouTube

When you start out as a fresher in the binary options trading industry, you Forex Historical Data Oanda must know Forex Historical Data Oanda all the ins & outs about this system. If you are not aware of the major terms and the overall process then, I would suggest you to follow this site: and go through the informative articles. He writes really good and highly informative articles about ... OANDA Rates® are calculated daily (Monday through Friday) and represent the previous 24 hour period aligned to UTC-midnight (8:00 PM Eastern Time). Bid, ask, and midpoint rates for the day are published and available no later than 10:00 PM Eastern Time. How to get Historical FX Data. Pick your base currency and the currency (or currencies) you want converted. Choose your timeframe (daily ... How Forex tick data can change your Forex vision for best. Forex Tester allows you to import an unlimited number of currency pairs and years of history data in almost any possible text format (ASCII *.csv, *.txt). We strongly recommend importing 1-minute data for accurate testing (it is possible to import higher timeframes but testing results may not be as good). Note: To increase the quality ... The Historical Forex Data service is openly available for all traders. We created this Historical Forex Data service to make it easier for the users of Forex Strategy Builder and Expert Advisor Studio to import good bar data. However, it is useful for all traders that need a reliable backtest in MetaTrader. Pop up a variety of customized graphs and cross-rate tables to analyze and compare historical exchange rates, over any time period since 1990. This is a Java-based application, so please be patient while it is loading. It uses OANDA's Foreign Exchange Protocol , which provides subscription access to OANDA’s currency data. FOREX: Foreign Exchange: MS7: To make things easier, here are a number of quick links for your daily downloads: Nov 06 2020: Nov 05 2020: Nov 04 2020: Nov 03 2020: Nov 02 2020: Oct 30 2020: Oct 29 2020 : Oct 28 2020: Oct 27 2020: Oct 26 2020: Oct 23 2020: Oct 22 2020: Oct 21 2020: Oct 20 2020: Oct 19 2020: Oct 16 2020: Oct 15 2020: Oct 14 2020: Oct 13 2020: Oct 12 2020: Oct 09 2020: Oct 08 ... Whether you need forex data for financial reconciliation, price localization, product development or more, we’ve got you covered. Leverage automated currency conversion directly into your app, website, ERP system, or accounting software. April 20, 2017. OANDA Solutions for Businesses. API, Best Practices, Announcements, Fintech. Tagged: FX, Exchange Rates, API, Tech, Fintech, Historical ...

[index] [6045] [14000] [17747] [25104] [5407] [17195] [10557] [9672] [15637] [27897]

Oanda - YouTube

Demonstrates how to easily acquire free historical data for your trading platform - in 3 simple steps! Note that this video has closed captions that can be translated into your local language. 30-50% profit as a US forex trader at OANDA. My US hedge forex trading strategy shows the way. - Duration: 32:00. Ryan Brown 13,153 views. 32:00. REST API concepts and examples - Duration: 8:53 ... History Get YouTube Premium Get YouTube TV Best of YouTube Music ... Discover answers to your questions and helpful information about the OANDA fxTrade forex trading platform. I added how to download Forex market data from OANDA in Python algo trading course with Crypto ... New MotiveWave 6 beta 1 testing best forex pairs now by Bryan Downing. 3:33 . We continue with ... You may not be seeing all of the Forex historical data that is available and that can be a bad thing. ★ Get clean, Daylight Savings Time adjusted MT4 data he... Oanda Company is a member of Forex international market and a reliable resource of currency rates. The Company has an access to the largest in the world quotation database with the most frequently ... In this video we are going learn how about the various sources for historical FOREX data. Primarily, we will be using data from Dukascopy bank. There are man...

https://binary-optiontrade.dialiothesurtels.gq