'The quantification of momentum in financial markets with RSI'

By David Sparks RSI was developed in 1978 by Welles Wilder and serves today as the key momentum oscillator widely used by traders the world over. Many traders use momentum oscillators (MO) as indicators of overbought (OB) and oversold (OS) conditions, however Wilder believed the divergence (DIV) between RSI and price action was a strong indication that a market turning point was imminent. The new price deceleration (PD) indicator quantifies the strength of an RSI divergence signal and is used to suggest the probability of the pending price reversal. This is based on the notion that RSI - PD values greater than 1.0 when spotted on a key technical level show the anticipation of support or resistance because if a level causes a strong momentum change more than likely it will also cause a price reversal. It won't work every time but the analysis of key levels in this manner yields high risk to reward ratios and provides a statist ical edge to the diligent trader. Caveat - currently this type of RSI signal has not been coded to flag PD values, and it must be done by hand but only takes moments of a trader s time. This simple axiom of relating high PD values to the strength of a psychological level is a way to qualify possible support or resistance for the purpose of buying into the weakness or selling into the rallies. Overall, the PD indicator is best applied outside the range when price tests a strong technical level, or the confluence of multiple levels. To further increase the probability of finding a valid price reversal, a top down fractal market analysis approach is used to seek multiple divergences in the popular time frames with PD values >1.0, this increases the edge of find a definitive reversal. Welles Wilder never intended the RSI indicator to be used primarily for its OB/OS alerts for trade signal generation; recent quantitative research supports the notion that these signals are not strongly indicative of price reversals. This is due to the fact that momentum oscillators like Stochastics and RSI can remain imbedded as price continues to trend. A very low or high print of the indicator does not relate to the probability of a price reversal. The use of OB/OS values is different than using the PD values of an RSI DIV signal because divergence reflect an aggressive slowing down of the current trend. Once a trend slows, it either; consolidates, retraces before continuing the trend, or makes a significant price reversal. Back testing shows that at important technical levels, strong and acute divergences are usually seen immediately prior to price reversals, therefore quantifying the divergence is very important. There are two types of divergences that momentum oscillators all produce, each is the opposite side of the coin for the technical trading of reversals. Wilder put forth the 'regular divergence' suggests a slow down of the existing trend, near the end of the move.

There are three types of regular signals that range from low to high intensity. The strongest of the signals is the classic divergence where price and RSI move exactly in the opposite direction. The second strongest signal is the double bottom / double top divergence where price remains flat and rsi moves opposite to it. These are strong reversal patterns when applied to market structure correctly. The weakest signal is the double bottom in the oscillator. The two strongest regular signals are seen below, while the oscillator double bottom has been omitted.

Fig 1. The two strongest regular RSI DIVsignals. The type 3, the weakest signal would be lower lows of price, like on the left, but the indicator would make a perfect double bottom.

Beyond regular divergence, Andrew Cardwell formulated the reverse divergence (rDIV) signal; his body of work explains how positive reversals are only found in up trends and negative reversals occur in downtrends. These signals are the opposite of regular divergence signals because instead of printing an RSI value where it should have been if momentum was constant, the rDIV signal is just a divergence print more than it should have been. The regular divergence is the opposite of this condition because an RSI divergence prints less than it should have been if momentum was constant. Explained another way, It's about higher highs and lower lows; If you find them in price, but not in the oscillator, you have regular divergence. If you find them in the oscillator, but not in price, then it's reverse divergence . The existence of the rDIV signal is used to confirm that a trend will continue, and is generally found at the third or more test of a trend line where price reversal tend to occur. These rDIV signals generally cannot be quantified with PD values because they often occur across a wider duration than regular RSI DIV. This means delta time is very high and makes the denominator large; so the PD values are less than 1.0. The rDIV signals can however be examined with respect to the change of RSI (dRSI%), and that will be explained next. The regular divergence, be it bullish (positive) or bearish (negative) is generally set at the start of a key market price reversal, and need not be found in a trend. More than often however, they are found on a key technical level, or at the third or more test of a trend line. The premise of this technical picture is logical; the regular divergence is a leading indicator' for a price reversal and it is most likely found at locations of strong support or resistance where price reversal are most probable. Therefore to find these signals outside the intra-day range where buyers and sellers are most under pressure then increases the odds of the reversal if it is seen on a key level. Out with the psychology of the market agents at extremes; price reversals need to be though of in respect to Newtonian behaviour. As with throwing a ball into the air and analysing its trajectory, the rate of change approaches zero and helps to pinpoints the top of the curve. The same principle is seen when price approaches an area of strong support or resistance, because divergences tend to be stronger when the market agents are anticipating the level holding. So how are divergence calculated? The RSI indicator line is calculated from price with a simple equation that you need not fully understand; all you need to appreciate is that the indicator line copies that of price with respect to where the peaks and valleys are printed in relation to one and other and compared to price. Meaning, in most cases, RSI mirrors that of price and when this is the case; price momentum is assumed to be constant. Put simply; if there are no divergence signals, then no assumptions about future price action can be made. The caveat here is that price reversals can occur without divergence; these are know as zero momentum price reversals and offer no predictive value in technical trading at market structural points.

In terms of convention; the start of the RSI DIV signal is called Time N , and the last RSI value, where the signal completes is know as Time N+1 . These points are just arbitrary points alone the x-axis that allow the calculate of the signal s delta time (change of time). The delta time (dT) is just the number of candles between the two points. In studying any chart, it is evident that 99% of the time, the RSI indicator peaks line up with the price peaks in the vertical y-axis. This is important because to calculate the delta RSI %, you need to use the cursor data from the charting package to read the RSI peak values that correspond to the price peaks in a classic DIV or a double top/bottom DIV. The following example from the S&P daily chart will make the calculation of PD clear. At the top or bottom of the range, RSI peak values are used, not RSI values from the valleys.

Fig 2. S&P daily chart up to June 4th, 2012. The high of 2012 sets a naked double top in price with a double bearish divergence; naked means it is not on a key level. At the top, the dRSI% equals -11.4% across 10 days; this gives it a PD = 1.1. The low on June 4th, is a doji, it is a reversal candle pattern, and positioned on a classic long level; the December high, the third test of a down trend line support and is 20 points below the 50% fib level from the current impulse wave. The classic bullish RSI DIV is the first signal after the high print which adds significance. The signal has a high predictive value for a price reversal because its found at some many key levels, has a high dRSI% at +12.6% and the PD value equals 1.3. The index rallied 200 points over the next three months.

In a classic divergence signal, price makes higher highs or lower lows while RSI forms the opposite pattern; the RSI values are then taken from the corresponding indicator peaks. The first RSI value is subtracted from the second, and if the absolute value is >5%, there tends to be a statistical edge that the signal suggests a pending price reversal. It is then assumed that larger dRSI% signals show a larger the drop of price momentum which then makes it more likely a price reversal is about to occur. This is only the case provided it occurs on a strong level, because naked signals have a better chance of being false signals. Overall, the dRSI% compares two points on the y-axis of the chart, and seeing the change of time is taken from the x-axis, it is then possible to contrast the two values to produce Price Deceleration (PD). The equation therefore becomes; PD = d RSI % / d time; put simply; price deceleration is equal to the change of RSI divided by the change of time. Strong PD values are important because the author holds the supposition that strong and acute divergences tend to precede price reversals if found on key technical levels. This indicator quantifies that behaviour. Quantifying RSI signals in this manner, should be familiar to anyone with knowledge of classical physics and the equation A = dV/dT. Acceleration (A) is just the rate of change of velocity, or the contrast of the x-axis to the y-axis, said another way, it is the ratio of velocity change divided by delta time. Seeing RSI is a momentum oscillator, the change of RSI is essentially a reflection of the market s velocity. It therefore stands to reason that the substitute of dRSI% into the acceleration equation is a logical step towards delineating momentum change in financial markets. The new equation for PD therefore quantifies the strength of the deceleration between two arbitrary points in time (N & N+1). Price deceleration values greater that 1.0 reflect the market agent s assumptions about how likely the level will hold. This is based on the assumption that near a key price level, price action slows down, an intense slowing suggests a greater psychological impact of that level. As seen in Figure 2, important locations with multiple confluence tend to have PD values greater than 1.0, but during times of headline risk, levels of little interest close to price action produce low PD values, so the market tends to penetrate them. Regardless of how strong an RSI DIV signal is, the indicator, like all indicator makes false signals. For example; in a 5min chart, if the market has sold off aggressively and reached the year low with confluence of the S1 pivot, if the first regular bullish RSI divergence appears and has a PD value >1.0, then the odds are good that the two levels may cause a retracement. Price deceleration values, when applied correctly, allow the trader a way to filter out false RSI signals and buy into the weakness or sell the strength with an edge.

When one looks back at support / resistance levels or key trend line tests that have held, it is noticed than momentum often changes as the level was touched. A close study of RSI signals shows there is a propensity for valid divergences that foreshadowed price reversals had a change greater than 5% (RSI is always between 0-100, so it is referred to as a percentage value). Sometimes key levels are penetrated by price, and a divergence is printed beyond the level; this can still mean a reversal because not all levels hold perfectly. When divergences are examined in double bottoms or double tops; the widest RSI peak values are used provided they correspond to candles with extremes that are exactly level (Fig 3). It doesn t matter if a one or two candle wicks are trimmed to select the signal, so long as the outside candles are on the same price level and correspond to clear RSI peaks (Fig 5). Trimming the rogue candle is not cherry picking the signal because the wick of one or two candles doesn't change the overall price pattern or the divergence. If a wick isn t trimmed, then valid double tops or bottoms can be missed. See the following two examples on double bottom or top divergences.

Fig 3. A standard double bottom divergence on the past low with a significant PD value. Notice it is 7 candles up and 8 candles down.

The example below shows the third test of a down trend line resistance where a reversal could be anticipated. If the two wicks were not trimmed then it would appear price made a lower low between the circle and the red arrow. In which case RSI dropped and matched price which is the normal pattern and would imply a momentum change. However, the green line trims 2 wicks and connects prices which are level; it therefore compares RSI values between two points which reveals a -11.9% bearish divergence. The trimmed double top is now a reversal pattern, and it is found at a point of price instability; the 3rd trend line test. It yields a good short entry point.

Fig 4. A trimmed double top divergence at the third test of a trend line resistance. In this example the first price peak (Time N) does not line up with the clear RSI peak, the rsi peak comes in the next candle. More than often, the first price peak lines up with the RSI extreme value. See the next chart for a better example.

Fig 4. GBPAUD 5min. A trimmed double top outside the trimmed wick therefore changes nothing.

The author purports that in a double bottom/top divergence pattern, the larger the dRSI%, the more probable a price reversal. As per the work of Gann, it also helps if the pattern is symmetric and has the same number of bars in each direction because this increases the probability that its a true reversal pattern. The 'perfect' double top or bottom is formulated as a pattern with PD = 1.0, clearly this is just a situation where the change of RSI between the peaks matches the number of bars in the signal. These types of divergences when found outside of the range on key levels are very strong reversal set ups. The principle of calculating a delta RSI% in the reverse divergence (rDIV) signal can also be applied. The larger the percent divergence; the higher the probability of a price reversal for trend continuation. As mentioned, reverse divergences are trend continuation signals, so the signal must complete at the third or more test of a trend line tag. In the higher time frames; the daily or hourly charts, these divergences tend to form across longer durations (dT) than regular DIV signals. This means the PD values of rDIV are often less than 1.0, and the calculation offers little predictive value because they are almost always less than 1.0 anyway. The PD of rDIV cannot be compared regular signals PD values, but if a rDIV signal is found with PD >1.0, then it more than likely very significant. It is for this reason that rDIV signals are best analysed with only respect to dRSI%. When reverse divergence is seen in the daily or hourly charts, more than often at the exact test of the trend line when the signal completes; a regular divergence signal will be found in the inferior chart. The regular divergence signal from the inferior chart when combined with reverse divergence from the higher time frame (superior) tends to confirm the key price reverse. The probability of a reversal will also be better if the rDIV has a dRSI% <5 and the inferior chart regular DIV signal has a PD value <1.0. See the following example in EURCHF recently, this strategy has been called RSI reloaded .

Fig 5. EURCHF hourly chart. At the third test of the new trend line support, the first rDIV signal appears. It has a large dRSI value at -18.8% which should be interrupted as a strong trend continuation signal. It formed across 26 hours (dT=26), so the PD value is <1.0 at 0.7. Was there regular RSI DIV in the 5min chart at Time N+1 where this signals completed? The 5min chart is next.

Fig 6. EURCHF 5min chart at the hourly chart s Time N+1 rDIV location. The first double bottom is naked (not on the key trend line support), however it has a high PD value. Hours later, when the trend line is tested exactly, the second double bottom divergence has a DIV greater than 5% which makes it significant, however the PD value is not significant because the signal takes longer to form than the amount of divergence. Signals with dRSI >5% on key levels with are not less valid as leading indicators for price reversals because PD is less than 1.0. High PD values only increase the odds that the signal is valid.

Overall, the calculation of PD is an improvement over dRSI% on its own because it quantifies the intensity of the momentum change. If the PD value is interrupted with respect to the market agent s expectations, it implies a greater disparity between the aggressive buyers and sellers at the level s test. If the move for the day sends price to the edge of its range to meet a strong technical level, then large PD values may imply the intra-day trend is about to reverse. Price Deceleration values are important and become more significant when applied to multiple time frames on strong technical levels. For example, an effective strategy to buy into weakness or sell into strength on the edges of a range is the 'High Momentum Price Reversal' (HMPR). This set up seeks a second regular RSI DIV signal in the lower time frame where the higher time frame signal completes (N+1). Lining up RSI DIV signals on key levels is a standard trading strategy, however this body of work improves on the set up by applying new principles to help filter out false RSI DIV signals. Therefore dRSI% and PD values, and seeking multiple divergences in different time frames increase the probability of finding the reversal point, a very difficult task for any trader. In summary; measuring the difference between the RSI peaks; to quantify the divergence becomes a way to filter high probability signals from lower probability ones. RSI signals greater than 5% tend to be more probable reversal indicators, it is also felt that larger dRSI% signals are more indicative of future price reversals. Moving on from this notion; price deceleration (PD), then quantifies how intense an RSI DIV signal is by comparing the DIV to its duration (dT). It is felt, higher PD values are more likely to precipitate price reversals when found on strong technical levels outside the day s range. Finally, combining the signal across multiple time frames also increases the probability of genuine momentum change that might foreshadow a price reversal. The following example demonstrates how PD is applied in technical trading. The author would like to thank Amplify Trading in London for their support, and their use of the strategy material in the f ollowing charts. Amplify trading specialise in professional education for traders, they use substantial trading experience to deliver intensive training to develop their new traders; helping them understand, analyse, and trade in today's financial markets .

Strategy

Entry-Long

Stop

1st Target

2nd Target

Neutral

141.24 S2

141.05

141.71

142.18

Amplify Trading Strategy, Oct 4th, 2012. The European economic calendar for the session ahead is light in terms of data. As such, safe haven bond markets are likely to focus on monetary policy developments from the BoE and the ECB, with no interest rate changes expected, and with the BoE expected to leave the APF unchanged at GBP 375bn. We are keen to monitor the Spanish yield curve as the ECB s press conference hits the wires. US Initial Jobless Claims (due to be released at 1:30PM BST) are expected to rise to 371k having fallen to 359k a fortnight ago. Factory Orders for August are due at 3PM, with the headline expected to decline from July s level of 2.8% to 2.6%. For today s session we prefer to maintain our bullish view on the Bund and would like to position ourselves with a neutral entry long at 141.24, the S2 pivot level.

Fig 7. Bund 30 minute chart. Notice the low from Oct 1st coming in as support on Oct 4th 2012. In this time frame, RSI copies price, so there are no signals to suggest evidence for a price reversal. When using the RSI Reversal Set up principle, the lower time frame is needed because RSI divergence there would suggest the level MIGHT hold as support.

Fig 8. Bund 5min chart up to the 13:30 candle. The bullish RSI divergence is +9.4% across 2 candles; the price deceleration (PD) value is 4.7 (PD = dRSI/dT). Comparing an RSI divergence across such a short time interval is not as valid, as when doing so across a slightly wider one. In this case, the 1 min chart should be examined there for an RSI signal, it will magnify the 5min signal and make it more clear. The S2 pivot at 141.24, is a location for a possible price reversal, so a momentum change here is very probable. The 1min chart is next.

Fig 9. Bund 1min chart. The last candle displayed is 13:34 exactly when the S2 pivot is tested. Leading up to then, RSI climbs from 13.8% to 35% across only 8 candles which makes the PD value = 2.7. Based on extensive research, price decelerations greater that 2.0 on technical l evels are strong indicators of price reversals, but did the Bund retrace?

Fig 10. Bund 1min chart. At 14:11 price tests the low from Oct 1st and as price makes lower lows, but RSI has climbed from 13.5% to 28.6% across only 5 candles. Price deceleration equals 3.0 and values >3.0 on technical levels are very strong indicators of price reversals when found outside the range. A strong PD value means that price has stopped it s advance, and this tends to precede price reversals, but not always. Therefore, the stronger the PD, the more probable a price reversal. What about the 5min chart at this point, was there also a bullish divergence in that time frame?

Fig 11. The 5 min chart Oct 4th 2012. The Bund has broken the Amplify neutral long entry by 11 ticks, but remains 9 ticks inside the stop level at 141.09. The RSI value has thrown its first bullish divergence in this time frame at +13.2% across only 9 candles. As mentioned prior; Price Deceleration contrasts delta RSI to delta Time and when the ratio is greater than 1.0, the signal tends to be a leading indicator of a price reversal. In this case PD = 1.5, and suggests a reversal might occur on the Oct 1st level. Seeing the level st from Oct 1 (141.13) caused bullish RSI divergence signals in the 1min and 5min charts, this makes it a high momentum price reversal (HMPR), as per the RSI Reversal Set up methodology.

Fig 12. Bund 5min chart. The bund bounced 60 ticks off the Oct 1st support level and the stop was not hit. The high momentum price reversal (HMPR) found dead on the Oct 1st low, quantifies that level as the most probable location for a price reversal. The Amplify Bund Strategy was perfect, they often select very good levels. While picking reversal points based on a fundamental and technical analysis is extremely difficult to do, the HMPR strategy can shed light on the entry in two respects. A) If this support level (141.13 Oct 1st low), was above the S2 level and a HMPR occurred before the entry was hit; it would suggest you could be a little more aggressive and get long before the lower entry level was tagged. This idea will help prevent missing moves B) The strategy entry was neutral long , which suggests position sizing can be reduced; if the HMPR is found on the entry level or very close to it but on a another key level; it suggests the entry level can safely be considered a LONG entry instead of a Neutral Long one. Overall, when a price reversal occurs with a HMPR, it quantifies that technical level as the strongest for the day. Last week, the Amplify US strategy perfectly three key levels that caused price reversal, and at each level, the HMPR was present. Pu t simply, Amplify perfectly picked those market tops and bottoms in EURUSD and OIL, and this is very difficult to do. The HMPR is a great was to confirm a valid entry.