Andrews Pitchfork: How to Draw It, the Schiff Variants and a Test of the Median Line

An Andrews pitchfork is a channel drawn from three swing points, labeled A, B and C. Its median line starts at A and passes through the midpoint between B and C; two parallel lines pass through B and C. The tool is named after Alan Hall Andrews, who taught it in a course on median line analysis. Traders use the fork to frame a trend: a move that stays inside the channel is read as a trend still in force, and the median line is watched as a possible target.

Who was Alan Andrews?

According to Optuma's history of median line analysis, Dr. Alan Hall Andrews studied engineering at the Massachusetts Institute of Technology and then at Harvard, and died in 1985; his date of birth is not known. He learned the idea of action and reaction, taken from Newton's third law, from the market analyst Roger Babson, and named his median line course the Action-Reaction Course in his mentor's honor. Optuma adds that Andrews gave his students' names to their additions: the Schiff adjusted median line comes from Jerome Schiff, a New York trader, and the Hagopian lines from another student.

How the pitchfork is built

Take three alternating swing points: for a rising fork, a low A, a high B and a higher low C; for a falling fork, the reverse. The median line runs from A through the midpoint of B and C and is extended to the right. The upper and lower lines are parallel to it and start at B and C. The variants differ only in where the median line starts, which TradingView's Auto Pitchfork help defines as follows:

Names are not used consistently. TradingView's own help page for its Schiff Pitchfork drawing tool puts the origin halfway between the first two points in both time and price, which its Auto Pitchfork page calls the modified Schiff. Our charts follow the Auto Pitchfork definitions. Moving the origin changes the slope of every line, and which variant is flatter depends on where the three points sit.

In the drawing, A sits well below B, so raising the origin to S or MS flattens the median line; with A close to B in price, the three lines would nearly coincide. Optuma, in its article on pitchfork methods, labels the three points PO (point of origin), P1 and P2 instead of A, B and C: the names change from one platform to the next, the construction does not.

How to draw a pitchfork, step by step

1. Find the swing that starts the move: a clear low for a rising fork, a clear high for a falling one. That is A.
2. Find the next swing in the other direction, the end of the first move: that is B.
3. Find the end of the correction that follows: that is C. In the textbook case it stays between A and B in price; in the Bitcoin example below it falls under A.
4. Draw the median line from A through the midpoint of B and C, and the two parallels through B and C. On TradingView the Pitchfork tool takes the three clicks; the Auto Pitchfork indicator picks the points itself.
5. Decide the variant before you look further right, and write down what would make the fork wrong, such as a close outside the lower line.

The points must be swing points, not any high or low. TradingView's Auto Pitchfork uses a “Depth” setting: with a depth of 10, each point must be the highest or lowest of the nearest 10 bars, five on each side. Our charts use the same kind of rule with 10 bars on each side, the equivalent of a depth of 20.

A pitchfork on daily Bitcoin

The chart below draws the fork by a stated rule: among all sets of three alternating swing points in the window, it keeps the one with the largest move from B to C. The points are A, a low on March 29, 2026 at 65,000; B, a high on May 6 at 82,850; and C, a low on June 5 at 59,130.91. C undercut A, so the fork rises but its lower line starts below A, a less tidy shape than the textbook one.

What followed, in full. On June 24, Bitcoin closed at 61,077.99, below the lower line, the kind of break that TradingView reads as a possible reversal. The fall went little further: the lowest close came on June 30 at 58,624.71 and the low of the move on July 1 at 57,800.19. Price then went back and forth across the lower line through July, closing inside the fork for a few days at a time (July 3 to 7, 9 to 12, 14 to 23 and July 26), stayed below it from July 27 to August 18, and closed back inside on August 19. It met the median line for the first time on September 21, the day of its high at 87,395.67, closed at or above it on September 21, 22 and October 4, and every other close to October 28 stayed inside the fork, below the median.

Does price return to the median line? A test of the 80% claim

Optuma's article on pitchfork methods, paraphrasing Andrews, says that when prices fail to return to the median line, “which studies have claimed to occur 80% of the time”, a reversal may be coming. It does not name the studies. We measured the claim on our own data.

Method: on the daily S&P 500 index from January 1970 to October 2026 and on daily Bitcoin from August 2017 to October 2026, we built an original fork on every three consecutive alternating swing points, with swing points of 5 and then 10 bars on each side. We started counting only once C was confirmed, 5 or 10 bars after it, and checked whether the median line fell inside a bar's range within the next 50 or 100 bars. To have a yardstick, we did the same for a control line: parallel to the median, on the other side of the C line and at the same distance from it, so outside the fork.

The median line was reached in roughly half of the forks, not 80%, and the control line outside the fork was reached more often in every row. The 50-bar chart counts 1,343, 665, 318 and 164 forks, row by row; the 100-bar table slightly fewer (1,337, 661, 311 and 162), because the most recent forks do not yet have 100 bars after them. In many cases the median had already been crossed before C was confirmed: 575 of the 1,343 S&P 500 forks with 5-bar swings. These figures describe the original fork drawn by one mechanical rule, without the judgment a chartist applies, and say nothing about trading results.

How traders use the fork

• Trend channel. TradingView's help says a trend is considered active as long as price stays inside the fork, and that reversals occur when price breaks out of it.
• Median line target. Prices inside the fork are watched for a move back to the median line. Our test above found no special pull.
• Halfway lines. Optuma describes internal lines drawn halfway between the median line and each parallel, used as intermediate levels inside the channel.
• Warning lines. Optuma describes extra parallels drawn outside the fork at the same spacing, used when price leaves the channel.
• Hagopian lines. Lines from the origin through B and C; Optuma presents a break of one of them, after price fails to reach the median, as a confirmation of a reversal.

None of these lines adds information: each is the same slope, or a line between two of the three points, moved by a fixed distance. They multiply the levels a price can be said to have respected, which makes a chart easier to explain after the fact and no easier to trade before it. Our control line in the test above is, in effect, a lower warning line, and it was touched more often than the median.

Limits of the Andrews pitchfork

• Choice of points. A different swing size or a different A gives a different fork.
• Choice of variant. On the Bitcoin example, the three variants met price on three different dates.
• The median line claim. In our data the median was reached about half the time, less often than a line outside the fork.
• No pattern test. Our bench tests chart patterns detected by fixed rules; channel patterns are covered on the channel pattern page.

For other tools that project lines from swing points, see the Gann fan, Fibonacci extensions and plain trendlines. More tools are in technical analysis.