FREE BOOK · LEVELS
Support, resistance and levels from A to Z: how well they really hold, measured on real candles
The Academy's tenth book in 14 chapters: what a level is and how it is tested, how to draw it, the previous-day levels, week, month and session levels, round numbers, how many touches, true and false breaks, the retest, trendlines and channels, confluence, entries, the stop and a system tested with cost and a random benchmark. Each chapter explains how it works, what we measured on gold and what to do with it. Mistakes, glossary and a final test.
14 chapters · 4 modules · 3 hours of reading, approx. · updated 2026-11-20
How to use this book
Every chapter has figures on real gold candles, four exercises with the solution explained and a recap of the key points. Do the exercises before you move on to the next chapter. Educational content, not financial advice.
Part 1: What a level is
What a level is, how it is tested and how it is drawn.
- 1What a level is and how it is testedA level is a price where the market is thought to react. We put it to the test: over 1,715 contacts with the highs and lows of the previous day, week, month and Asia the price "held" in 48% of cases and broke in 51%, like random levels shifted by a few ATR (49%). Levels are not useless, but they hold like a coin. · 13 min
- 2How to draw a levelA level is drawn with a price (the high or low of a swing) and a zone around it. We tried three timeframes and three thresholds: swing levels hold 48-49% (random levels 47-49%), and the width of the zone changes the number of contacts but not the outcome. There is always a level nearby: within 1 ATR of the price there are on average 2.8, and a 50 $ round number in 74% of candles. · 11 min
Part 2: Where levels go
Previous-day, week, month and session levels, round numbers and touches.
- 3The previous-day levelsYesterday's high, low, close and open are the most used levels. On gold they hold the same (48% over 830 contacts, random levels 47-49%). The next day touches yesterday's high or low in 50% of cases and the close in 84%; the hold changes with the hour (57% in the American morning, 45% in the evening), but it changes almost the same way for random levels too: it is the clock, not the level. · 11 min
- 4Week, month and session levelsThe previous week's, previous month's and Asia levels do not behave better than the others: over 1,715 contacts with time-based levels the hold is 48% (± 2), like random levels (47-51%). By type (high 54%, low 39%, close 56%, open 43% for the week) the values swing inside the error (± 7-13), and the month's levels are 20 in ten months: too few. The hold changes hour by hour, but it changes for random levels too. · 10 min
- 5Round numbersMultiples of 50 and 100 $ hold 44%, a little less than grids of non-round prices (46-51%, mean 49%): a weak effect, which with ten grids to compare is almost explained by chance. They do not attract the extremes: 20% of the daily highs and lows fall within 5 $ of a multiple of 50, exactly what chance expects. · 11 min
- 6How many touchesA level touched several times holds neither more nor less: on time-based levels the hold is 47%, 50%, 48%, 50% and 55% (± 11) from the first touch to the sixth and beyond; after a hold the next touch holds 49%, after a break 51%. Neither the time elapsed nor the type of level changes the result: the price does not "remember" touches. · 9 min
Part 3: What happens to levels
True and false breaks, the retest, trendlines and channels and confluence.
- 7True and false breaksWith one close beyond the zone the price comes back (a false break) in 42% of cases on time-based levels; asking for two closes it falls to 31%, at 0.5 ATR to 26%, at 1 ATR to 12%. The same numbers come out with random levels: the frequency of false breaks depends on how much you ask for to say there is a break, not on the level. · 9 min
- 8The retest after the breakAfter a true break the price comes back to touch the old level from the new side in 75% of cases within three days (median: 10 candles). The "support that becomes resistance" holds 50% (± 4), like random levels (50%). A resistance broken upwards: the retest holds 56% (± 6); a support broken downwards: 45% (± 6): differences inside the error. · 10 min
- 9Trendlines and channelsTrendlines (lines through two consecutive swings) hold 47-48% (random levels 46-48%); the "third touch" holds less (42%), but random lines at first contact give 45% too. The channel's parallels hold 50% (random 47-50%). Deleting the line at the first break drops the hold to 41%, and that of random lines too (44%). · 10 min
- 10Confluence85% of contacts have another level of a different type within half an ATR, and random levels have 86%: confluence is the norm. With more overlapping levels the hold does not rise (50%, 48%, 47%, 46% with 0, 1, 2, 3 or more other levels, like random ones). A level near a round number holds 44%, against 51% for the others. The best of 19 pairs (Asia plus trendline: 64%) is within the range of chance. · 10 min
Part 4: Trading and the system
Entries, stops, a system on trial, mistakes, glossary and final test.
- 11Entries at levelsNine entry rules at levels (bounce, break and retest, on time-based levels, H4 swings and round numbers), with stop, target at 1 R, costs and one trade at a time, give mean R between −0.17 and +0.06, with errors of ± 0.08-0.15. None beats random direction on the same entries (−0.02 to −0.04). The bounce at round numbers (−0.17 ± 0.09) is the only one that stands out, on the negative side. · 11 min
- 12Where to put the stopA level that holds is still pushed through: in 50% of holds the price goes beyond by at least a quarter of an ATR before turning, in 11% by at least 1 ATR (median 0.26 ATR). A stop half an ATR beyond the zone is hit in 76% of contacts, one at 1 ATR in 67%. The return after the stop, the "stop hunt", happens in 45% of cases (at 0.5 ATR), like random levels (47%). Size changes with the stop: to risk 100 $, 5.5 ounces with 0.25 ATR and 3.4 with 1 ATR. · 11 min
- 13A levels system on trialFour variants built from what the chapters found (bounce at time-based levels, without round numbers, with confluence, break of H4 swings) give mean R from −0.05 to +0.03 (errors of ± 0.08-0.10 over 367-552 trades). The best is exceeded by 32% of trials with random levels; chosen on the first half, the break of H4 swings is worth −0.03 ± 0.14 in the second. To measure R with ± 0.05 about 1,500 trades would be needed, that is two to three years. · 10 min
- 14A complete day, mistakes, reference, glossary and final testA real day (17 March 2026) with all the levels and six trades out of six on target, read next to the ten-month average (−0.03 R per trade), the most common mistakes in chapter order, the quick reference of rules and measurements, the glossary of terms and a 15-question final test. · 19 min