I would start by breaking your code into the top of the box and the bottom of the box. Place each of these in the relevant exploration columns: {ColA} {Breakout of Darvas Top} NewH:=ValueWhen(1,HIGH>Ref(HHV(HIGH,5),-1),HIGH); box1:=HHV(HIGH,3)<HHV(HIGH,4); BoxTop:=ValueWhen(1,BarsSince(HIGH>Ref(HHV(HIGH,5),-1))=3 AND box1=true,NewH); x:=C>BoxTop; x AND Alert(x=0,2); {ColB} {Falling through Darvas Bottom} LowL:=If(LOW=LLV(LOW,5),LOW,If(Ref(LOW,-1)=LLV(LOW,5),Ref(LOW,-1),If(Ref(LOW,-2)=LLV(LOW,5),Ref(LOW,-2),If(Ref(LOW,-3)=LLV(LOW,5),Ref(LOW,-3), If(Ref(LOW,-4)=LLV(LOW,5),Ref(LOW,-4),0))))); box1:=HHV(HIGH,3)<HHV(HIGH,4); BoxBottom:=ValueWhen(1,BarsSince(HIGH>Ref(HHV(HIGH,5),-1))=3 AND box1=true,LowL); x:=C<BoxBottom; x AND Alert(x=0,2); Now all you have to do is filter for a 1 in Column A (the close went above the top of the box today, but was below the top of the box yesterday) or a 1 in Column B (the close is below the bottom of the box today, but was above the bottom of the box yesterday). {Filter} ColA>0 OR ColB>0 Hope this helps. wabbit :D

BTW - the Darvas Box formulas can be simplified somewhat.... --8<---------------------------- Hc:=BarsSince(H>Ref(HHV(H,5),-1))=3 AND HHV(H,3)<HHV(H,4); BoxTop:=ValueWhen(1,Hc,HHV(H,5)); {plot} BoxTop; --8<---------------------------- and --8<---------------------------- Hc:=BarsSince(H>Ref(HHV(H,5),-1))=3 AND HHV(H,3)<HHV(H,4); BoxBottom:=ValueWhen(1,Hc,LLV(L,5)); {plot} BoxBottom; --8<---------------------------- so the whole lot together makes, if I am not mistaken: --8<---------------------------- {Darvas Boxes} {recoded by wabbit 06/06/06 - just gotta love this date!} Hc:=BarsSince(H>Ref(HHV(H,5),-1))=3 AND HHV(H,3)<HHV(H,4); BoxTop:=ValueWhen(1,Hc,HHV(H,5)); BoxBottom:=ValueWhen(1,Hc,LLV(L,5)); {plot} BoxTop; BoxBottom; --8<---------------------------- I am pretty sure that they could be simplified further, but I didn't play with them for too long, and I don't want to! wabbit :D P.S. NEVER take someone else's code at face value. ALWAYS read and understand each and every line to see if it makes sense.

Benny, Uncheck "load minimum records" and try again with at least 200 records loaded! Have a read through the forum for all the posts about this feature, and how many times it has brought people and explorations, unstuck. wabbit :D

I suppose I should offer some more by ways of an explanation of what is happening... If I break a part the code for the boxes into: {ColA} {Darvas Boxes - Hi side breakout} Hc:=BarsSince(H>Ref(HHV(H,5),-1))=3 AND HHV(H,3)<HHV(H,4); BoxTop:=ValueWhen(1,Hc,HHV(H,5)); x:=C>BoxTop; x AND Alert(x=0,2); {ColB} {Darvas Boxes - Lo side breakout} Hc:=BarsSince(H>Ref(HHV(H,5),-1))=3 AND HHV(H,3)<HHV(H,4); BoxBottom:=ValueWhen(1,Hc,LLV(L,5)); x:=C<BoxBottom; x AND Alert(x=0,2); {Filter} ColA>0 OR ColB>0 In each condition, there is no 'initialisation' so for the filter to return a value BOTH Column A AND Column B criteria must return a value. If in the number of bars loaded in the exploration, Column A never happens OR Column B never happens then the result will be a NA rejection. To ensure the filter is accurate, you need to include enough history so that at least once in the loaded data, the close has been above the top of the box AND the close has been below the bottom of the box. The exploration will never return stocks where the close has been continuously inside the box, or has only ever escaped outside the box on one side only. If you want, you or someone else, could add an itialisation code. Roy Larsen is the expert at this, so check out his work in his latches code and apply the same principles. Or you can just make sure you have enough data loaded! Hope this helps. wabbit :D P.S. Please make sure you fully understand the Boolean operators, AND, OR, NOT etc They are extremely important in the post above and throughout MS.

Hi Mazzyon. as I understand the Darvas system, it goes on the attack when a stock rises above its one-year high. having previously tried to break thru on several occasions, and forming 'boxes' along the way. something like that. this would make your last question to Mr Wabbit very plausible.

myself, writing exploration the oldfashioned way, have always used the following line in my Darvas searches, in conjunction with various other criteria:

CLOSE>Ref(HHV(CLOSE,120),-1)

this will give me a break on today's close, thru the High for the last 6 months. I think mr Darvas would agree that things move faster these days & to ask for a 1-year high is excessive. anyway, the 120 can be changed to 250 which is one year's trading days.

perhaps you could ask mr Wabbit would he be kind enough to fit this into the formula somehow, then we shall both benefit. you could also include something with regard to minimum volume, in order to cut out a lot of low-liquidity junk.

P.S. as far as the Darvas system is concerned, I don't believe it made provision for the downside. except to the extent that mr Darvas worked to very strict Stoploss guidelines. that would have been in part at least, the secret of his success. maybe he kept an eye on a watch-list of potential break-thru candidates, in that case the formula as it stands, would be sufficient.

Amory has most of the situation covered wrt Darvas. If you want to only include the instances when the stock is making its all time high (for the data loaded) add the Highest() function to the filter condition; e.g

{Filter}
ColA>0 OR ColB>0
AND C=Highest(C);

or you could do the timeframe lookback like Amory has already displayed.


Hope this helps.

wabbit [:D]

as usual, spot-on Jose! on chart, indicates not just any old entries, but specifically when they fit into a Darvas pattern. it's a top indicator alright.

when signaling 'exit', could mean part of a new box-formation leading to better things, or just that: exit, get out, it didn't work! I do believe old Darvas worked with a very tight stop-loss, something like 2%. no benefit of the doubt there!

thanks Wabbit and Jose, for clearing that one up.