Bickford throws away Elliott's subjective wave counting and replaces it with a mechanical swing-reversal algorithm and a lettered taxonomy of two, three, four, five, and six wave cycles. The forecasting tables built on top of that taxonomy explain, by the book's own math, barely a third of the next wave's size.
The one idea
The book's actual idea has nothing to do with Elliott's five-wave count. Bickford builds a 'swing chart' from raw price ticks using two settings: a minimum fluctuation unit (one pip) and a minimum reversal amount (how many pips of retracement it takes to plot a new diagonal). Every continuous move becomes one straight-line 'wave,' and a run of waves is a 'cycle.' Because the rule is purely mechanical (a reversal is a reversal once it clears the threshold), two people running the same data through the same reversal amount get identical wave counts. That is the entire pitch: objectivity Elliott Wave never had.
On top of that mechanical chart, Bickford classifies every three-wave cycle into nine named shapes (impulse, rectangle, contracting and expanding triangles, connector) purely by comparing the heights of adjacent waves with greater than, less than, or equal to. He then runs ordinary least squares regression against a EURUSD 2005 tick database (just under 8 million ticks) to see whether the height of wave 1 and wave 2 predicts the height of wave 3. The basic model comes back weak: an average coefficient of correlation around 31 percent across reversal amounts from 5 to 50 pips. Splitting the sample into terciles (wave 2 under one third of wave 1, between one third and two thirds, or over two thirds) sharpens the fit in the first tercile to the 78 to 88 percent range, but that tercile is also the smallest sample, down to a few hundred waves at wider reversal amounts.
The back third of the book (chapters 27 through 29) turns the same swing data into a fractal tool: 'cycle collapsing' merges qualifying three-wave cycles into a single wave using channel-line coordinate rules, so a trader can step up to the next larger wave size without recomputing from raw ticks. It is a genuinely clever bit of chart mechanics. It is also the only part of the book that ships with a worked, repeatable procedure rather than a probability table.
What it actually teaches
Bickford's forecasting process, reconstructed from the chapters that carry it, runs in this order.
SET THE SWING PARAMETERS
Pick a minimum fluctuation unit (one pip for most pairs) and a minimum reversal amount (the book tests 5, 10, 15, 20, 25, 30, 35, 40, 45, and 50 pips). The reversal amount is how far price must retrace before the algorithm plots a new wave instead of extending the current one. Smaller reversal amounts generate far more waves: at 5 pips the EURUSD 2005 sample produced 912,385 waves, at 50 pips only 538.
BUILD THE SWING CHART
Run the raw OHLC or tick data through the reversal algorithm in chapter 9 (initialize price and time arrays, track direction, extend or reverse the current wave based on whether the move exceeds the box size times the reversal amount). The output is a series of diagonal waves connecting peaks and valleys, with lateral chop filtered out entirely.
MEASURE THE LAST TWO WAVES AND PICK A TERCILE
Compare the height of the most recent wave to the one before it. Test A: wave 2 is less than one third the height of wave 1. Test B: wave 2 is between one third and two thirds of wave 1. Test C: wave 2 is over two thirds of wave 1. Each tercile has its own regression table (17-1, 17-2, 17-3) with different A and B coefficients and a different coefficient of correlation (COR).
APPLY THE REGRESSION FORMULA
Wave 3 = (A x wave 1) + (B x wave 2), using the A and B values from the tercile table that matches your reversal amount. Worked example from the book: a 10-pip reversal amount, wave 1 = 19 pips, wave 2 = 12 pips. Since 12 falls in the second tercile (between one third and two thirds of 19), use A = 0.1474, B = 0.9843: wave 3 = (0.1474 x 19) + (0.9843 x 12) = 2.80 + 11.81 = 14.61, rounded to a 15-pip forecast, with the table's COR of 73.70 percent cited as the confidence level for that reversal amount and tercile.
CROSS-CHECK WITH THE MEAN RATIO SHORTCUT
A faster, less precise version: multiply wave 1 by the tercile's mean ratio (0.68 for the same example) to get roughly the same 13-pip estimate. The book offers this as a mental-math substitute for running the full regression.
COLLAPSE CYCLES TO STEP UP A FRACTAL LEVEL
If the wave pattern qualifies (start point of wave 1 lower than or equal to the start of wave 3, end point of wave 3 higher than or equal to the end of wave 1, for a bull cycle), draw a single line from the cycle's first point to its last and treat the whole three-wave cycle as one wave in the next larger fractal chart. The book reports this reduces a 39-wave chart to 13 diagonals in one worked example.
What it looks like Monday morning
A reader who takes this seriously would stop staring at candlestick patterns and instead build a swing-reversal chart at a fixed pip threshold, then log the height of every two-wave and three-wave sequence against what the next wave actually did. That is a legitimate, testable habit and the book hands you the reversal algorithm in full pseudocode to do it.
What a Monday-morning reader would not get from this book is a trade. There is no entry trigger tied to the wave forecast, no stop placement, no position sizing, and no account for spread or slippage anywhere in the method chapters. The forecast tables output a probable wave height, not a signal to buy or sell, so the actual trading system has to be built entirely by the reader on top of Bickford's statistics.
| Model | Sample size | A coefficient | B coefficient | Coefficient of correlation |
|---|---|---|---|---|
| Basic model, all waves (Ch. 16, Table 16-2) | 61,961 | 0.3452 | 0.6960 | 34.12% |
| Second-tercile model only (Ch. 17, Table 17-2) | 5,602 | 0.1474 | 0.9843 | 73.70% |
The book's worked example (wave 1 = 19 pips, wave 2 = 12 pips, forecast wave 3 = 15 pips) uses the second row; the improvement in correlation comes from narrowing the sample to one tercile, not from a better underlying model.
The generic model above therefore is inadequate as a valid third-wave forecaster.James L. Bickford, Forex Wave Theory
Where it fails
- The whole statistical base is one currency pair, one year. Every regression table, frequency table, and worked example in the forecasting chapters comes from a single EURUSD tick database covering 2005 (with practical studies pulled from January through March 2006). The book claims the tables 'can be applied to the other major currency pairs as well... with only minor modifications' but never shows that test. A method validated on one pair in one calm, pre-crisis year is not validated.
- The author's own numbers show the base model barely works. Chapter 16's autoregressive model, wave 3 forecast from wave 1 and wave 2, returns a coefficient of correlation averaging 31 percent across ten reversal amounts. Bickford states plainly that this means less than one third of the third wave's height can be attributed to the two prior waves, and calls the model 'inadequate.' The tercile split in chapter 17 helps, but the strongest results (COR near 80 to 88 percent) come from the first-tercile subsample, which shrinks to as few as 68 waves at a 25-pip reversal amount, a sample size where a clean-looking correlation is easy to get by chance.
- No entries, exits, stops, or position sizing anywhere. The method produces a probabilistic estimate of the next wave's height and a percentage confidence figure, then stops. There is no rule connecting that forecast to a buy or sell order, no stop-loss placement, and no discussion of position size or risk per trade. A reader has to design the entire trading system around these tables from scratch.
- Transaction costs are never mentioned. The swing data is built from raw mid-price ticks with reversal amounts as tight as 5 pips. At 5-pip thresholds the book's own frequency table shows near-even odds of the third wave running the same or a different direction, margins that a typical retail spread would eat into before the forecast even resolves. Spread, slippage, and commission do not appear once in the 241 pages.
- Roughly a quarter of the book is reference padding. Parts 3 and 4 (chapters 7 through 15, about 65 pages) walk through point and figure charts, Renko charts, Gann angles, the Kondratiev wave, and Elliott and Gartley history before the book's own method even starts in chapter 16. The appendices add ISO currency pair codes, exchange rate tables, and global banking hours, none of which touch the forecasting method.
- Never revisited since 2007, and the market regime moved on hard. The tables reflect EURUSD behavior from 2005 and early 2006, a period of relatively contained, trending dollar weakness. Nothing in the book accounts for the 2008 crisis, the 2015 Swiss franc peg break, the negative-rate era, or the volatility spikes of 2020 to 2022, and there has been no updated edition to re-run the regressions against that data.
Who it is for
Buy it if
You want a rigorously mechanical, non-discretionary way to draw swing charts and classify wave patterns, and you are willing to treat the regression tables as a starting hypothesis to backtest yourself across more pairs and years than the book covers. Programmers who want the swing-reversal algorithm in ready-to-code pseudocode will get real value from chapters 9 and 27.
Skip it if
You want a trading system with entries, exits, and risk rules, or you are looking for evidence this approach makes money in live or even properly out-of-sample backtested conditions. Traders hoping this is a modern, computable version of Elliott Wave that finally works should also skip it: the book's own correlation numbers are the strongest argument against using it as-is.
As an Amazon Associate I earn from qualifying purchases. It costs you nothing extra, and it does not change what gets recommended or what the verdict says.