End of Quarter - March 30, 2012

[Billy]

A 20-day look at the PM Money Flow shows a most interesting long term reversal pattern with lots of positive divergences and confirmed support by the average MF. Plenty of indicators derived from price alone obviously do just what stock prices do and are matching the existing trend with a lag time. Users of this website prefer to trust the leading indicator of Money Flow and to follow the edges derived from backtesting of its past behavior. This can be puzzling as MF is often too early but is usually right in the end with a high percentage of wins. Losses are few and limited by numerous fail-safe protection triggers in case of trouble.

13656
13658

The GDX robot came 9 cents away from hitting the stop (47.94) before a strong reversal both in price and MF in afternoon trading. After being minutes away from a short signal, it is now at 3 average days of issuing such a signal, giving us a lofty cushion for the last day of the first quarter. Floor clusters are neutral and don’t matter much as they will be totally different next Monday with new weekly, monthly and quarterly levels. We’ll have plenty of time to peruse them quietly this weekend.
Billy

13657
13655

[TraderD]

This can be puzzling as MF is often too early but is usually right in the end with a high percentage of wins.


Billy, running the risk of repeating earlier comments I made, my interpretation of the MF/PA (price action) relationship is different and points to a mismatch in timeframes rather than MF being merely "early" relative to PA. The GDX chart is "technically broken" with selloffs commencing right at exhaustion points of recoveries (from oversold conditions). This is consistent with the longer timeframe as captured by the declining average MF line. Not taking this into account is, in my opinion, the reason for the poorly-timed RT robot entries and the stark discrepancy between the MF/RT curve and corresponding PA when comparing the two charts you posted. I'm somewhat relieved by the change of tone in the past few months from seeing RT as a major breakthrough to a more skeptical analysis of its dynamics relative to the EOD robot.

$.02,

Trader D

[Billy]

Billy, running the risk of repeating earlier comments I made, my interpretation of the MF/PA (price action) relationship is different and points to a mismatch in timeframes rather than MF being merely "early" relative to PA. The GDX chart is "technically broken" with selloffs commencing right at exhaustion points of recoveries (from oversold conditions). This is consistent with the longer timeframe as captured by the declining average MF line. Not taking this into account is, in my opinion, the reason for the poorly-timed RT robot entries and the stark discrepancy between the MF/RT curve and corresponding PA when comparing the two charts you posted. I'm somewhat relieved by the change of tone in the past few months from seeing RT as a major breakthrough to a more skeptical analysis of its dynamics relative to the EOD robot.

$.02,

Trader D

Doron,
We are very open-minded. What back-tested algorithm do you propose to add to the EOD robot to take into account timeframes mismatches under all market conditions? We offer you a $.02 subscription if you come with one.
Billy

[TraderD]

Doron,
We are very open-minded. What back-tested algorithm do you propose to add to the EOD robot to take into account timeframes mismatches under all market conditions? We offer you a $.02 subscription if you come with one.
Billy

That's a rather compelling offer, Billy, with a very clear payoff :-)

My first inclination would be to re-examine the hard-coded threshold governing the basic calculation of the MF and comparing backtest runs with different settings of those thresholds to see if the results are consistent across the range of values. There aren't too many of them, which is the good news. I would start from the 20-day span for MF calculation, e.g. use 5-day, 10-day, 40-day, 60-day, 90-day, etc. Based on the outcome of this basic test, there are other ideas to pursue which relate to multi-timeframe rule sets (may cost more than $.02)...

Trader D

[Billy]

That's a rather compelling offer, Billy, with a very clear payoff :-)

My first inclination would be to re-examine the hard-coded threshold governing the basic calculation of the MF and comparing backtest runs with different settings of those thresholds to see if the results are consistent across the range of values. There aren't too many of them, which is the good news. I would start from the 20-day span for MF calculation, e.g. use 5-day, 10-day, 40-day, 60-day, 90-day, etc. Based on the outcome of this basic test, there are other ideas to pursue which relate to multi-timeframe rule sets (may cost more than $.02)...

Trader D

These tests have already been made and 20 days was the most consistent outcome. That’s why it has been chosen.
Billy