The technique which he found to be the best (among many which he shows producing winning results) is called the G3M1 method. This is allowing a winning bet to ride three times (parlay) and doubling any losing bet one time only (a one stage Martingale). According to the author this takes a 100 unit bankroll and the gambler must be patient as he cycles through his negative/net gain excursions. He provides quite a few tables to support his contentions about this and other similar combinations.

While I haven't even experimented with any of these methods, this book is interesting reading. I personally don't find that kind of "grind" or up and down form of play to be very interesting. But, I have run into a few people doing the Koetsch methods. Some claim to have similar results as the author. I don't have any substantiation of the claims independent of what is provided in the book, but I will say that the book is well written and the data is convincingly presented."

The Koetsch Method is played on the even money chances only on one side Player/Banker; Red/Black etc.. Someone posted the following instructions in another forum:

How to Bet:   Start at one unit.
if it wins let it ride 3 times and pull out. If you lose along the way start over.
If any 1 unit wager loses double once (if that won pull 2 units off leaving 2 for the next bet in the win seq.)

An all win sequence will look like this: 1,2,4,8,(Win, Ride1, Ride2, Ride3)
An all lose sequence will look like this: 1,2,1,2,1,2,1,2 (Lose, Dbl lose)

Playing this way it will win on any streak of 4 or more and also win on a streak of 2 or less (If this one wins treat it as a win of 1 and let the one unit plus the one originally bet ride). The important part of the system is to not play all the way though the 100 decisions and instead stop when the stop win is reached. You will see in the majority of 100 round sessions that you can quit ahead a quite few times. This one will usually pull to the plus side 3 times per 100 dec. on average. This means if it swings south 3 times it usually will return 3 times.
Remember to wait till the block of 100 is over to start again or move to another table upon winning. If you never reach your win goal just stop positive or play to the end. It will almost always be less of a loss than the peak loss per session since even bets usually tend to balance. This one is straight up math and no voodoo involved.
You guys that can test this stuff easily, please give it a go. You might like it. You can also tweak the times it rides or the double ups to produce different results but at the cost of more risk or less reward. This one is easy to bet on both sides of the fence too (black and red etc.) and bet the difference. "

You look at the last 2 decisions. If it chops, you bet the opposite of the last decision to continue with the chop. If it repeats, you bet the same as the last decision to continue with the streak. If it goes 2-2, you follow 2-2 until it breaks.

Here is a comparison between flat betting and using the G3M1.

A round is 100 decisions and 600 rounds (60,000) decisions were tested for each of the following examples in the Koetsch book.

Flat betting...

[attachimg=1]


G3M1....

[attachimg=2]


There are some interesting findings such as the percentage of times you will be ahead by 20 units flat betting = 2% compared to 47% using the G3M1. Also the percentage of times you will be ahead by 30 units flat betting = 0% compared to 29% using the G3M1. Now you may say OK, it's not that surprising that you aren't going to get ahead by 20-30 units flat betting over 100 decisions but then you need to look at how the average loss per round is 1.38 units flat betting compared to 3.96 units using G3M1. So if you go to 'gamble' and try and maximize your winnings / potential winnings, you can see why Koetsch prefers the G3M1.

Thanks Bally

It is not something I've gone into depth with,  I do recall there was a certain amount of interest in G3M1 some time back.

Kudos for posting that, saves us the trouble.

Thanks, Lugi for posing this, I have the book :-)

Cheers

G3 is 1,2,4 and not 1,2,4,8.