Relative Strength Index (RSI): A Case Study (how to use/not use it)
We're going to start from scratch and first understand what RSI measures and then move on to using statistical data from the @BrighterData API for BTC and 12 alts for the past 8 years for findings and useful frameworks.
- What does RSI really measure?
The relative strength index is basically a scoreboard of the last 14 candles i.e. out of all the movement, how much of it was up?
An RSI of 50 means its a tie. Subsequently, an RSI of 70 means the up move has been beating the down move by ~2.3 to 1 lately.
Note how there is nothing being calculated regarding the absolute price being "too high" or what the fair value of an asset should be i.e. when a market is grinding up the RSI will just stay high because it is a trending regime and the same goes for the other way around.
Across 25,000 trading days, here is what the daily RSI looks like in each market mode (P.S. just used a simple heuristic of using price above the 20d EMA as an uptrend, sideways for anything else and downtrend being below 20d EMA).
Note how in an uptrend an RSI of 70 is basically a normal day (happens 20% of the time)
Finding 1: An elevated RSI especially on a higher timeframe is a green light framework.
When a sideways market prints an elevated RSI (atleast by the time it gets to the daily) it's usually breaking out & sign of momentum which carried through another 7% in the next days and 14% in the next 21 days i.e. preemptively attempting to fade where it feels good because coin is "overbought" is usually the worst spot to do so.
Finding 2: "Buy Oversold" does not work unless there's a floor
The only place "oversold" might POTENTIALLY pay is when there is a sideways market or consolidation that is establishing a floor. Subsequently if you're waiting for RSI to have a deep correction in an uptrend you're basically waiting for a structural break which will end up being a regime shift.
Finding 3: The deeper the dip, the worser the outcome
The typical number that is used is a reset into 40-50 before buying a trend pullback.
A strong asset or a trending regime typically will not offer a steep discount, by the time it does there is a regime shift and it's actually showing damage.
A continuation framework can be a shallow dip on H4 in a daily uptrend, this is one of the most well-behaved frameworks that was found during this study.
Finding 4: Using solely RSI, you can see momentum building but buying early doesn't offer any "edge"
"Earliness" isn't free when using the RSI as the sole framework. By the time it offers any meaningful statistically proven return on price extension, the move looks "telegraphed" but the conclusion really is just that the market has picked up steam in one direction and will continue in that direction.
Finding 5 - Alt Specific - Check BTC mode before reading any alt signal
Alt ignition with BTC trending is the real trade.
Alt ignition with BTC chopping - you're front running BTC breakout with extra steps or essentially expressing with increased leverage that BTC will continue to trend higher.
A lot of the alt repricing from the lows can be attributed to a mechanical trade due to the sheer fact that they were heavily sold off. A touch of optimism along with price up = more capital gets allocated = higher price and the loop continues until new information enters the market.
There have been outliers that have been evident during this cycle where some assets continue to get repriced based on the sheer fact that they are revenue generating and/or the promise being sold & price being the best narrative continues to carry the upward drift trajectory (fundamentals).
This leads into my next point of one of the best ways use to use RSI as a screener i.e. ranking alts against each other to determine which vehicle is best to hold when the high beta basket is repricing higher.
Finding 6 - A Useful Case for RSI is a ruler for Comparison
The way RSI is calculated it has one property that no test could eradicate and that isn't a claim about the future because it is bounded, unitless and relative in nature it fits in perfectly as a use case to compare different assets at the same time.
What better way to use a RELATIVE strength index than to use it for a RELATIVE comparison?
You can create your own screener but one useful indicator in TV is: RSI Screener by Leviathan
That's cute and all but how do I use this?
In a trending regime, you can use the h4 trend as a precursor. When BTC dips, with one quick scan determine which alts are retaining their momentum the best by looking at this view.
You add some confluence using your current execution framework - is the dip into some key area, is it retesting a ltf trend which retaining the htf structural integrity? etc.
READ THIS IF YOU SKIPPED EVERYTHING ELSE.
So what did we learn?
-
Throw "overbought" and "oversold" out. Determine what regime you're in and understand that elevated RSI is just a relative term without any context. In other words, in a tug of war if one side is winning you wouldn't automatically assume the other team will start winning immediately? (analogy for automatically thinking about fading a move)
-
Use relative strength for relative comparison especially in a trending environment (there are other ways of doing this such as a regular price spaghetti chart but RSI will normalize this over 14 periods), this will provide you with a quick scan of where strength lies.
-
The H4 momentum reset framework combined with daily trend is a somewhat sound principle to investigate. Ideally you want to watch for alts that don't do harsh pullbacks (~60 RSI cooldown ideal) when BTC dips while it is in a trending regime.
-
There is a lot of dispersion so understand that longing a bunch of dog shit just means you're essentially just carrying a lot of leveraged beta risk on a wider mechanical trade of everything getting dragged up vs. being highly selective on demonstrated strength combined with some tailwind or proven metrics (some assets in this cycle have been carried by a narrative tailwind, fundamental backed tailwind or a combination of both).
