RSI(2) and 9 Days at Extremes
The S&P 500 has traded 9 days in a row at an extreme RSI(2) reading over 90. Unfamiliar with RSI(2)? Read more here and here.
Echoing Rob Hanna at Quantifiable Edges, extended overbought readings like this were much more likely in the last century when the market was momentum-driven (i.e. up days tended to beget more up days), but are a rarity in today’s mean-reverting market (read more).
Prior to this month, this century had only seen one instance of a 7-day RSI(2) run over 90 – we’re now at 9-days and counting.
Where does the market go from here?
I’m hedging my bets.
The bearish side of me says that by almost any quantitative measure, probability favors a pullback. But hell, I would have said the same thing most of last week.
The cautious side of me says that this market is temporarily being driven by completely non-normal forces. Our abnormal market filter (AMF) stands at 78% and we’ve been slowly pairing down position sizes in our YK and Scotty strategies since July 15th as a result – I do not want to want to be picking up nickels in front of a bulldozer with a drunk man at the wheel.
The graph below shows the abnormal market filter (red) vs the S&P 500 (grey) MTD. As the filter creeps up, we ratchet down exposure (note: the AMF is available daily in the detail box of the free State of the Market report):
There is no bullish side of me – I think long and strong at this moment is just begging for a roundhouse kick to the face.
The net effect is YK and Scotty will be a bit short on Monday, but not enough to hurt should this market keep on keepin’ on.
What about RH and the original MS strategies? The original MS strategies are long-only and don’t trade even remotely overbought markets – they’ve been in cash most of the month. RH is a product of our Timer Seeds program, and because I’m not the developer, it doesn’t make use of the AMF. Yes, it’s been fighting the trend, and yes, it’s getting butchered this month.
One last note, because of the nature of how RSI(2) is calculated, basically any close down from here takes us back below 90. To be specific, if the S&P 500 closes below 978.46 on Monday, we’re under the threshold. 90 isn’t a magic number, just a reference point, so treat it for what it’s worth.
Happy Trading,
ms
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Random Stuff | 9 Comments
this link http://www.insidercapital.com/BullishReview/COTINDEX.HTM summarizes the actions of commercial traders (known through CFTC docs to be goldman sachs, the fed and JP Morgan primarily) in the futures and option index contracts vs institutional players (hedge funds and mutual funds) and small traders broken down into percentiles. Prior to the early july downturn the commercials sold the bulk of their positions and were near zero, on July 7 prior to the recovery they were at 100%, last week they cashed in to 84% and now they are at 100% . This explains the major buying occuring early last week. I have found this useful in understanding the market manipulations and forces at work behind what the indicators tell us. All of my fundamental knowledge and research suggests that the Fed and Goldman are attempting to manipulate the market higher to prevent a major fallout and buy time—this also makes logical sense. That said if you look on WSJ online they have a separate portion detailing money flows——-selling on strength and buying on weakness. Whenever the S&P500 is in the top 5 the market sells off soon after as it reflects the actions of the “big money.” Curiously Goldman Sachs was in the top 5 late last week as well–coincidence? It seems to me that the big boys have been unloading their positions into this rally. A pullback seems logical at some point this week. Keep an eye on this stuff—–it seems to offer a unique perspective on the forces that can be bigger than the weight of empirical evidence.
cheers
dv
dv:
This is interesting, but also pretty confusing.
Would you provide a link to the WSJ Online money flows data to which you are referring?
Also, would you explain your comments a little more?
Thanks,
Walter
Michael, re: rsi2 > 90 for numerous days..
I don’t have access to my research, but after x days, it actually becomes bullish for RSI2 to be extended above 90 for multiple days.
Of course we will expect some weakness 1-3 days out, but fairly quickly after that the edge gets bullish again, except for one factor, and that is win/loss ratio, which mostly favors the bears.
My RSI2 system uses an ab filter similar to yours. I highly recommend some sort of filter like this to protect from losses that may come from shorting into an RSI2 that stays above 90 for multiple days.
As an aside, it might be that once RSI2 > 90, and doesn’t go back beneath for x days, is a long signal.
RE to Wood: hello sir – saw that analysis on your blog and I think that makes sense for folks holding longer positions. In my own case though I’m never looking more than one day out though (and potentially moving long to short and back again daily) so I’m still looking at it as (one of many) very short-term bearish signs. michael
Absolutely understand. That is how I tend to use it as well, but I was surprised at what it was showing 7-20 days out.
the link to wsj money flows is http://online.wsj.com/mdc/public/page/2_3022-mflppg-moneyflow-20090723.html?mod=mdc_pastcalendar
Thank you dv for the links. I reckon it is a remarkable insight. Proving your intuition would require the historical data of COTIndex.htm. I may save the pages every day from now on. Or just follow it. Unfortunately, gathering if for 1-2 months is not statistically significant data. But at the extreme, this can even be a base of another bullish/bearish ‘state of the market’ indicator if proved right…. and if Michael likes it as well. :)
“gathering if for 1-2 months” -> “gathering it for 1-2 months” (sorry for the typo)
Michael -
You’ve raised an excellent point regarding how to interpret extreme RSI readings.
As you pointed out, there’s a lack of data to study these extreme readings –
“Prior to this month, this century had only seen one instance of a 7-day RSI(2) run over 90 – we’re now at 9-days and counting.”
When you look at RSI(2) extreme readings at a smaller, intra-day time frame (where extreme readings of RSI92) = 99+ happen almost every several days), you’ll see that at some point, RSI(2) stops behaving as an oscillator and appears to become a measure of the strength of the trend. Often in the intra-day time frame, a RSI(2) reading of 99+ indicates a trend that’s going to continue for at least another 60 minutes. Over the course of the next 60 minutes, the RSI(2) may drop below 90, but it seldom drops below 50, forming a series of peaks with modest valleys. It doesn’t drop below 10 until the trend has died out (something that doesn’t always happen on an intraday period).
It would be interesting to measure the length/strength of an extreme RSI(2) reading and see what is average trend length after RSI(2) drops below 90 for the first time.
Thanks again for the excellent series on RSI(2)
-F.M.