There are a number of ways to calculate daily pivots (those R1, S1, R2, S2, etc.). For example, if you go here, you can see Classic, Woodie, Camarilla, and DeMark pivots. These pivots have fixed formulas, typically based on the open, high, low, close of the prior day. Since they are based on rigid mathematical formulas, I cannot understand why they would work. I don’t see any logical reason why the market will adhere to levels calculated using fixed mathematical formulas, especially when there are so many different subjective mathematical formulas put forth by different people.
Regardless, I am always interested to read research on this to learn more on the validity of pivots, and found a few pages below. The studies below also covered 3 different markets (futures, forex, stocks).
In general, no strong conclusions here. In terms of study methodology though, Bulkowski’s seem the most robust.
- The first one is here, which concludes that pivot points marginally work.
- For each day, the study calculated the percentage of area between S3 and R3 covered by the 7 pivot bands (S1, S2, S3, P, R1, R2, R3, each band is + and – 0.1% of each pivot level). The E-mini futures were used.
- Over the study period, say the average daily percentage was 30%.
- Then it assumed that if pivots work, then the pivots must be good for identifying reversals, meaning that the High or Low of the day will occur within a pivot band. And to be considered “good” for identifying reversals means that the percentage of time the High or Low occurs within a pivot band is more than 30% (or whatever the percentage is).
- This makes the assumption that if pivot points don’t work, then the probability of the High or Low price landing in a pivot band should be the same as the area covered.
- I think that reasoning is not really correct. When you show that the probability of the High or Low price landing on a pivot band is higher than the area covered by pivot bands, it just shows that the High and Low prices are not uniformly distributed across S3 and R3. This reflects on the nature of the High and Low prices, and not on the nature of the pivot points.
- One way to normalize this is have a case of some equally spaced reference pivots between S3 and R3 to make the comparison, rather than simply using the area percentage.
- A 630-day period is also too limited for a study because whether the market is in an uptrend or downtrend can reflect on the positioning of the daily highs and lows within S3 and R3.
- The second study is here, which concludes that pivot points have no significance at all.
- The study was done on the forex market using 10 currency pairs.
- It compared the trading results of a strategy where you enter immediately upon detection of a strong uptrend / downtrend, versus waiting for retracement to a pivot point before entering.
- The results show that there is no statistical difference.
- I find this study pertains more to the probability of a retracement during a strong uptrend / downtrend rather than whether prices pause or reverses at pivot points.
- Thomas Bulkowski also did some research here, concluding that pivot points work (50-60%)
- He tested using 55 stocks using the 1-min timeframe. For each pivot point, he calculated the percentage of times the pivot point acted as support / resistance. If prices sailed through 40% of the time but turned 60% of the time, then the pivot points worked better than expected (i.e. 50%).
- Most pivot points had scores of 50-60%, which was better than expected.
- Linda Raschke Bradford found no statistical significance in pivot points
- “Everything I do is based on actual chart points. I’m always looking at the swing highs or the swing lows. I never calculate Fibonacci numbers, Gann retracements, artificial pivot points or other things like that because I’ve never found any edge or any statistical significance from testing them.
- But I can quantify chart points. I can quantify and test something like, ‘If the market made new momentum lows and there’s a reaction up by half an ATR, what are the odds the market will trade below that low?’ I can determine there’s, say a 68-percent probability of that happening.”
Shown is an interesting YouTube video by Adam Grimes (SMB Capital) where he showed that if you simply draw random levels while hiding the price chart, when you put the price chart back on, you will easily find a number of levels looking significant. This experiment also shows how easily it is to be persuaded / misled that there is significance in pure randomness.