Why Smart Traders Watch the First Hour (Initial Balance Guide)

Yup, that is from Friday, February 27th.

Dark Mode!

Here are today's trades. All shorts. Decided to switch to the dark mode on charts, Expert Advisor and terminal. I see that 33 people have downloaded the indicator I made ( https://www.forexfactory.com/thread/...5#post15571535 ), I'd like to know how it's working out for you guys that downloaded and any feed back.

Yay..dark mode rocks!
Save your precious eyesight

Okay everyone, here's the session break out indicator for dark mode. Please let me know what y'all think about it!

Just in case here are the instructions:

There are some instructions to it:

1. If there is a session you don't want to see, just set the start and end times to zero (00:00)
2. "Session for BO" is the session that you want to measure for a break out. (So let's say you want to measure the U.S. session initial balance, you enter 16:30 in "Session_for_BO_Start" and 17:29 in "Session_for_BO_End")
3. IMPORTANT the "Session for BO End" needs to be 1 minute before the session end time. So for example the initial balance is the first hour of the trading session right? Let's use the U.S. as an example. That would be technically from 16:30 to 17:30. BUT, by giving MQL5 the time stamp of 17:30 the code will wait for the 17:30 candle to close and measure the whole high and low of that candle. So if you want to correctly measure the high and low of the first hour the "Session for BO End" will need to be 17:29.

Here is an example of what the indicator should look like:

The session boxes are visible without being distracting. I like how the breakout level is clearly defined after the initial range. Have you tested how it behaves during high-impact news inside that measured session? Does it sometimes give false breaks if volatility spikes right at the session close?

During high impact news it lags a tiny bit to draw the box and levels. It updates with each tick. As far as volatility right after the session close? Yeah you'll get the alert. Also, thanks for downloading it and for the questions!

'Morning everyone! Took some losses today but all under control. Faked out but managed to recoup part of the losses. Done for the day.

A lot of uncertainty in the markets and that reflected in this week's PnL. Just a few bucks away from braking even for the week but I believe I'm just done for the week since tomorrow is NFP.

Most likely we will see very little trends until the mid-term elections in the U.S. I do believe macro-trends do affect this strategy even if it's a day trading strategy.

I will be back to Python to work on tweaking this strategy.

I am traveling..Monday trade was shown live (see the link below), here are results this week, had to withstand moderate DD(in circles), but trusted my Money Management and controlled my MindFuck(the urge to get out of the trade), +9.5%

https://www.forexfactory.com/thread/...5#post15595875

Hey everyone! I've been running statistics with Python on opening ranges. Using 5-minute candle data exported from MT5, we're measuring 15 minute opening ranges. Here are the python results:

Date range: 2011-06-14 to 2026-03-06
Total days analyzed: 2621
========== STAT 1: OR SIZE vs CONTINUATION RATE ==========
OR Size Days Continued Rate Avg Pts Cont
0-5 430 335 77.9% 6.44
5-10 846 668 79.0% 12.45
10-15 572 454 79.4% 21.23
15-20 294 241 82.0% 27.59
20-25 196 164 83.7% 28.14
25-30 129 106 82.2% 36.92
30-40 105 80 76.2% 39.71
40-50 26 17 65.4% 60.36
50+ 23 17 73.9% 67.35
===========================================================
========== STAT 2: HOW OFTEN DOES PRICE TAG OPPOSITE SIDE FIRST? ==========
Days price tagged opposite side first : 1037 / 2621 (39.6%)
Days price did NOT tag opposite side : 1584 / 2621 (60.4%)
Avg pts into opposite side (when tagged) : 17.11 pts
Median pts into opposite side : 9.50 pts
Max pts into opposite side : 342.58 pts
When price TAGGED opposite side first → continuation rate : 49.0%
When price did NOT tag opposite side → continuation rate : 99.4%
=============================================================================
Bullish OR — tagged opposite side : 513 / 1354 (37.9%) | Avg pts into opposite: 19.15
Bearish OR — tagged opposite side : 524 / 1267 (41.4%) | Avg pts into opposite: 15.11
========== STAT 3: WHAT HOUR DOES CONTINUATION HAPPEN? ==========
Hour Count % Cumulative%
17:00 1632 78.4% 78.4%
18:00 185 8.9% 87.3%
19:00 99 4.8% 92.0%
20:00 68 3.3% 95.3%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
23:00 1 0.0% 100.0%
Total days with continuation : 2082
==================================================================
========== STAT 4: OR SIZE vs POINTS CONTINUED ==========
OR Size Days Cont Avg Pts Cont Median Pts Avg OR Ratio
0-5 335 6.44 4.20 3.70 1.74x
5-10 668 12.45 8.18 7.30 1.71x
18:00 185 8.9% 87.3%
19:00 99 4.8% 92.0%
20:00 68 3.3% 95.3%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
23:00 1 0.0% 100.0%
Total days with continuation : 2082
==================================================================
18:00 185 8.9% 87.3%
19:00 99 4.8% 92.0%
20:00 68 3.3% 95.3%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
18:00 185 8.9% 87.3%
19:00 99 4.8% 92.0%
20:00 68 3.3% 95.3%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
23:00 1 0.0% 100.0%
19:00 99 4.8% 92.0%
20:00 68 3.3% 95.3%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
23:00 1 0.0% 100.0%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
23:00 1 0.0% 100.0%
23:00 1 0.0% 100.0%
Total days with continuation : 2082
==================================================================
==================================================================
========== STAT 4: OR SIZE vs POINTS CONTINUED ==========
OR Size Days Cont Avg Pts Cont Median Pts Avg OR Ratio
OR Size Days Cont Avg Pts Cont Median Pts Avg OR Ratio
0-5 335 6.44 4.20 3.70 1.74x
5-10 668 12.45 8.18 7.30 1.71x
10-15 454 21.23 17.87 12.30 1.73x
15-20 241 27.59 22.50 17.39 1.59x
20-25 164 28.14 21.48 22.28 1.26x
25-30 106 36.92 30.10 27.19 1.36x
30-40 80 39.71 34.95 33.34 1.19x
40-50 17 60.36 35.55 45.03 1.34x
50+ 17 67.35 40.10 60.19 1.12x
==========================================================
Charts saved to: C:\Users\eramirez_OSF\Desktop\orb_extended_stats_charts.png
Full statistics saved to: C:\Users\eramirez_OSF\Desktop\orb_statistics_extended.csv
PS C:\Users\eramirez_OSF> ^C
PS C:\Users\eramirez_OSF> & "C:/Program Files/Python314/python.exe" c:/Users/eramirez_OSF/Documents/ORB_Statistics.py
Data loaded successfully. Rows: 649570
Date range: 2011-06-14 to 2026-03-06
Total days analyzed: 2621
============================================================
========== STAT 2: OPENING RANGE SIZE ==========
Average OR Range : 12.92 pts
Median OR Range : 10.25 pts
Smallest OR : 0.61 pts
Largest OR : 87.25 pts
10th Percentile : 4.07 pts
25th Percentile : 6.10 pts
75th Percentile : 16.75 pts
90th Percentile : 25.75 pts
========== STAT 3: POINTS MOVED ABOVE / BELOW OR ==========
Avg pts moved ABOVE OR High : 13.21 pts
Avg pts moved BELOW OR Low : 15.20 pts
Median pts above OR High : 6.50 pts
Median pts below OR Low : 6.40 pts
Max pts above OR High : 444.02 pts
Max pts below OR Low : 289.15 pts
========== STATS 4 & 5: BULLISH OPENING RANGE ==========
Total bullish OR days : 1354
Days price closed above OR High : 1099 (81.2%)
Days price did NOT continue up : 255 (18.8%)
Avg pts continued up — all bullish days : 14.11 pts
Avg pts continued up — only when continued : 17.38 pts
Max pts continued up : 421.65 pts
========== STATS 6 & 7: BEARISH OPENING RANGE ==========
Total bearish OR days : 1267
Days price closed below OR Low : 983 (77.6%)
Days price did NOT continue down : 284 (22.4%)
Avg pts continued down — all bearish days : 17.00 pts
Avg pts continued down — only when continued : 21.91 pts
Max pts continued down : 287.52 pts
Overall continuation rate : 2082/2621 (79.4%)
Overall reversal rate : 539/2621 (20.6%)
========== DAILY CLOSE vs OR LEVEL ==========
Bullish OR — daily close ABOVE OR High : 669 (49.4%) | Avg: 16.92 pts | Median: 12.05 pts
Bearish OR — daily close BELOW OR Low : 530 (41.8%) | Avg: 21.85 pts | Median: 12.68 pts
========== STAT 8: OR SIZE vs CONTINUATION RATE ==========
OR Size Days Continued Rate Avg Pts Cont
0-5 430 335 77.9% 6.44
5-10 846 668 79.0% 12.45
10-15 572 454 79.4% 21.23
15-20 294 241 82.0% 27.59
20-25 196 164 83.7% 28.14
25-30 129 106 82.2% 36.92
30-40 105 80 76.2% 39.71
40-50 26 17 65.4% 60.36
50+ 23 17 73.9% 67.35
========== STAT 9: OPPOSITE SIDE TAGGING ==========
Days price tagged opposite side first : 1037 / 2621 (39.6%)
Days price did NOT tag opposite side : 1584 / 2621 (60.4%)
Avg pts into opposite side (when tagged) : 17.11 pts
Median pts into opposite side : 9.50 pts
Max pts into opposite side : 342.58 pts
When tagged opposite → continuation rate : 49.0%
When NOT tagged → continuation rate : 99.4%
Bullish OR tagged opposite : 513/1354 (37.9%) | Avg pts: 19.15
Bearish OR tagged opposite : 524/1267 (41.4%) | Avg pts: 15.11
========== STAT 10: HOUR OF CONTINUATION ==========
Hour Count % Cumulative%
17:00 1632 78.4% 78.4%
18:00 185 8.9% 87.3%
19:00 99 4.8% 92.0%
20:00 68 3.3% 95.3%
21:00 56 2.7% 98.0%
22:00 41 2.0% 100.0%
23:00 1 0.0% 100.0%
Total days with continuation: 2082
========== STAT 11: OR SIZE vs POINTS CONTINUED ==========
OR Size Days Cont Avg Pts Cont Median Pts Avg OR Ratio
0-5 335 6.44 4.20 3.70 1.74x
5-10 668 12.45 8.18 7.30 1.71x
10-15 454 21.23 17.87 12.30 1.73x
15-20 241 27.59 22.50 17.39 1.59x
20-25 164 28.14 21.48 22.28 1.26x
25-30 106 36.92 30.10 27.19 1.36x
30-40 80 39.71 34.95 33.34 1.19x
40-50 17 60.36 35.55 45.03 1.34x
50+ 17 67.35 40.10 60.19 1.12x

What do these numbers mean?

The opening range between 16:30 and 16:45 averages about 13 points wide and acts as a very reliable directional signal. When the OR closes bullish, price goes on to close a candle above the OR high 81% of the time during that session, and when it closes bearish, price closes below the OR low 78% of the time.

On average price travels about 13-22 points beyond the OR boundary when it does continue, and roughly 1.7 times the OR size for smaller ranges. Almost all of that movement happens fast — 78% of continuation days see the breakout happen right in the first hour at 17:00, and by 18:00 you've captured 87% of all continuation days.

The sweet spot for OR size is between 15-25 points where continuation rates hit 82-84%, while very small ORs under 5 points and very large ORs over 40 points are slightly less reliable.

The single most important finding is the opposite side tagging stat. On 40% of all days, price briefly dips below the OR low on a bullish day or spikes above the OR high on a bearish day before eventually continuing. When that happens, the continuation rate drops to just 49% — essentially a coin flip.

But when price does NOT tag the opposite side at all, the continuation rate is a staggering 99.4%. This means the OR boundary on the wrong side is an extremely meaningful level — if your stop loss sits just beyond it and gets hit, the trade is almost certainly failing.

It also means the best trades are the clean ones where price immediately moves in the OR direction without any whipsaw.

Great stuff eadweardcap,

Not a lot of people here are applying python to find patterns in the market. And all of these findings are extremely impressive and practical.

I only want to say one thing: When finding patterns like this, the very first thing you need to do is test for statistical significance. This is important, because even if the pattern looks consistent, we have to understand the randomness of this phenomenon. If it's statistically significant, you have a much higher confidence to trade this pattern or turn it into some kind of a strategy. Assuming your analysis was done by AI (seems like it by the way the output is strucutred) I would suggest discussing the topic of "statistical significance" in the context of patterns with your AI. You will understand how deeply important this process is before coming to any conclusions.

I urge you to keep going.

Just curious as to whether you python results adjust for USA daylight savings, which has just occured.

Nice work, however, because of your hard work, I will point you to a biggest mistake that you and 95% of other "want to be traders" that do not succeed are making. It is that, entries do not mater, it is how you manage a trade to your full TP.

Hundreds of people stopped at your thread, just to playa hate people who is making money and 95% of them go and continue looking for nonexistent holy grail indicator for entries.

The reason I never opened my own thread is because no one wants to work hard and pay attention to what is important in trading, which is MM and controlling MindFuck(Fear and Greed).

I am reposting my Weekly equity chart(from above), as 2 circles represent -3.5% and -5% DD(-$3,500 and -$5,000), but because of the believe in my MM and ability to control MindFuck, I was able to see 9.5% profit. It represents $9,500 per each 100K.

Took 10 years (5 years of hard work to realize what I written above) and 5 years of trading live(100%+). Rinse and Repeat.

https://www.forexfactory.com/thread/...7#post15603157

I'll get on that right now. I've ran like a trading strategy with it and asked it to do Montecarlo simulations. I'll share that later.

Unfortunately no. It's one of the down sides of back testing this strategy. The sorta good thing is that when the time change occurs it's only one week that the hours are mismatched.

Ok, so building a strategy around the statistics from the previous post (https://www.forexfactory.com/thread/...1#post15603811) I ran a montecarlo simulation and did the statistical significance that @amvt mentioned. Here's the results:

Montecarlo:
Running 10,000 simulations of 668 trades each...
Simulations complete.
========== MONTE CARLO RESULTS ==========
Simulations run : 10,000
Trades per simulation : 668
--- FINAL PnL DISTRIBUTION ---
Actual historical PnL : 1391.09 pts
Average final PnL : 1394.94 pts
Median final PnL : 1396.06 pts
Best case (99th pct) : 2269.88 pts
Good case (75th pct) : 1651.85 pts
Median (50th pct) : 1396.06 pts
Bad case (25th pct) : 1140.63 pts
Worst case (1st pct) : 512.95 pts
--- DRAWDOWN DISTRIBUTION ---
Actual historical DD : 111.68 pts
Average max drawdown : 191.99 pts
Median max drawdown : 180.48 pts
Worst drawdown (99th) : 380.00 pts
Worst drawdown (95th) : 307.01 pts
Worst drawdown (75th) : 222.31 pts
--- PROBABILITY OF PROFITABILITY ---
Profitable simulations : 9,998 / 10,000
Probability profitable : 100.0%
Probability of loss : 0.0%
==========================================

Attached Image (click to enlarge)

Statistical Significance Results:
Total trades: 668
=================================================================
SECTION 1: OVERALL STRATEGY SIGNIFICANCE
=================================================================
Question: Is the overall 57.6% win rate real or random luck?
Overall strategy
Trades : 668
Wins : 385 (57.6%)
p-value : 0.0000
95% CI : 54.4% — 100.0%
Result : HIGHLY SIGNIFICANT
Meaning : Less than 1% chance this is random luck.
=================================================================
SECTION 2: SIGNIFICANCE BY YEAR
=================================================================
Question: Which individual years show a real edge?
2016
Trades : 15
Wins : 11 (73.3%)
p-value : 0.0592
95% CI : 48.9% — 100.0%
Result : BORDERLINE
Meaning : Could be real but not enough evidence yet.
2017
Trades : 3
Wins : 0 (0.0%)
p-value : 1.0000
95% CI : 0.0% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
2018
Trades : 58
Wins : 31 (53.4%)
p-value : 0.3470
95% CI : 41.9% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
2019
Trades : 44
Wins : 31 (70.5%)
p-value : 0.0048
95% CI : 57.2% — 100.0%
Result : HIGHLY SIGNIFICANT
Meaning : Less than 1% chance this is random luck.
2020
Trades : 92
Wins : 55 (59.8%)
p-value : 0.0379
95% CI : 50.7% — 100.0%
Result : SIGNIFICANT
Meaning : Less than 5% chance this is random luck.
2021
Trades : 93
Wins : 50 (53.8%)
p-value : 0.2670
95% CI : 44.7% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
2022
Trades : 63
Wins : 36 (57.1%)
p-value : 0.1568
95% CI : 46.0% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
2023
Trades : 113
Wins : 62 (54.9%)
p-value : 0.1734
95% CI : 46.7% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
2024
Trades : 86
Wins : 50 (58.1%)
p-value : 0.0803
95% CI : 48.7% — 100.0%
Result : BORDERLINE
Meaning : Could be real but not enough evidence yet.
2025
Trades : 90
Wins : 52 (57.8%)
p-value : 0.0851
95% CI : 48.6% — 100.0%
Result : BORDERLINE
Meaning : Could be real but not enough evidence yet.
2026
Trades : 11
Wins : 7 (63.6%)
p-value : 0.2744
95% CI : 35.0% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
=================================================================
SECTION 3: SIGNIFICANCE BY DAY OF WEEK
=================================================================
Question: Is Friday's 67% win rate real? What about other days?
Monday
Trades : 129
Wins : 80 (62.0%)
p-value : 0.0040
95% CI : 54.4% — 100.0%
Result : HIGHLY SIGNIFICANT
Meaning : Less than 1% chance this is random luck.
Tuesday
Trades : 142
Wins : 71 (50.0%)
p-value : 0.5334
95% CI : 42.8% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
Wednesday
Trades : 139
Wins : 76 (54.7%)
p-value : 0.1544
95% CI : 47.3% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
Thursday
Trades : 130
Wins : 72 (55.4%)
p-value : 0.1271
95% CI : 47.8% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
Friday
Trades : 128
Wins : 86 (67.2%)
p-value : 0.0001
95% CI : 59.7% — 100.0%
Result : HIGHLY SIGNIFICANT
Meaning : Less than 1% chance this is random luck.
=================================================================
SECTION 4: SIGNIFICANCE BY OR SIZE BUCKET
=================================================================
Question: Is the 15-20pt OR bucket genuinely better?
OR 5-10 pts
Trades : 22
Wins : 10 (45.5%)
p-value : 0.7383
95% CI : 27.1% — 100.0%
Result : NOT SIGNIFICANT
Meaning : Could easily be random chance. Treat with caution.
OR 10-15 pts
Trades : 420
Wins : 237 (56.4%)
p-value : 0.0048
95% CI : 52.3% — 100.0%
Result : HIGHLY SIGNIFICANT
Meaning : Less than 1% chance this is random luck.
OR 15-20 pts
Trades : 226
Wins : 138 (61.1%)
p-value : 0.0005
95% CI : 55.4% — 100.0%
Result : HIGHLY SIGNIFICANT
Meaning : Less than 1% chance this is random luck.
=================================================================
SECTION 5: ADDITIONAL EDGE METRICS
=================================================================
Profit Factor : 1.399 (above 1.5 is good, above 2.0 is excellent)
Expectancy : 2.082 pts per trade
Average Win : 12.68 pts
Average Loss : -12.33 pts
Win/Loss Ratio : 1.028
Kelly Criterion : 16.4% (suggested max risk per trade)
Profit Factor verdict : BORDERLINE — marginal edge, monitor closely
Kelly verdict : Positive edge confirmed. Suggest risking 16.4% or less per trade.
=================================================================
SECTION 6: WIN/LOSS STREAK ANALYSIS
=================================================================
Question: How bad can losing streaks realistically get?
Max consecutive wins : 10
Max consecutive losses : 6
Avg win streak length : 2.4
Avg loss streak length : 1.7
Statistically expected max loss streak : 7.6 trades
Actual max loss streak : 6 trades
Verdict : NORMAL — within expected range

Yeah daylight saving is a right pain in the ass.
Two solutions. If u can get the data with US timestamps then no daylight saving needed for the US indices. Also if CSV has a volume column there's a very obvious spike every day at 9.30 EST. You can run the CSV through Python to check that the 9.30 opening time is correct , and if there is a few weeks in the year it's not (by volume surge) then adjust the data for these days +1 or -1 hours and resave the CSV.

just in case, this small util will should align summer/winter clock shifts with the rest of dataset. Unless you have some exotic time format there

Absolutely Amazing. I am honestly more impressed by your persistence to keep going since this stuff gets extremely technically and almost everybody will trade the pattern rather than stress-testing it.

Having said that, there are some extremely peculiar results and I would like to shed light on these observations:


This stat tells me right away that something extremely bad happened with the technique you were using or the results you are analyzing. Because probability of loss can NEVER be 0. Even the "best" strategies will have tail-risk and this stat is extremely non-realistic.



You ran 10k simulations. Assuming it was vanilla Monte-Carlo and you were not utilizing GPU, this would've taken you a lot of time. If this took a few minutes instead of hours, understand that something is wrong with simulation. Also, it's always a better idea to have a larger sample of trades. 668 is nowhere near enough for statistical significance. Although, this topic can be a bit tricky.


Again, this stat is also extremely non-realistic. a p-value of less than 0.0000 means that's almost guaranteed. This usually never happens and I suggest you let the AI walkthrough your code to understand what's happening because the numbers are suspicious.

I would advise you to consider walk-forward permutation test. This is a well-documented technique to gauge statistical significance. If you want to guide the AI and also for personal use I would recommend Timothy Masters, he's a brilliant author and has great books on backtesting that will give you the knowledge and wisdom to tackle this properly.

Otherwise, if the results are truly genuine (which is highly unlikely from my experience) these are some of the most impressive stats I have ever seen in my 4 years of Quantitative Finance career! Even beating many academic paper's stats. Hence, I highly suggest going through this once again, but with the proper knowledge. This stuff is never easy. But the hard part is always starting it. And you are the only one I have seen on FF that is doing this and getting the well-deserved recognition. I suggest keep going and learning this. Eagerly waiting for the results

Did AI write your post? (obviously)
Eh? Takes milliseconds. Probably plot rendering took longer