The best way to do it would be to wipe out the US carrier forces in the first six months - say, Big-E/Lex/Sara at Pearl, followed by Hornet, Yorktown, Wasp and maybe Ranger at alt-Coral Sea/Midway. Both battles turned on sufficiently fine margins that it wouldn't take much to turn them, and without breaking JN-25, Midway would be utterly transformed. This would give them at least twelve months carte blanche before the first of the war-built carriers started arriving.
I see the key implications of this as follows:
1) The Japanese will be able to continue to extend their hold over the south-west Pacific, and give them more time to fortify them, forcing the counter-offensive to reclaim more of these outer islands, and probably at even greater cost than OTL.
2) Far stronger forces will be available to respond to the counter-offensive, in terms of both carriers (no OTL Midway, more time to carry on building) and aircrew (no Santa Cruz, more time to carry on training).
3) The counter-offensive will begin against a backdrop of prolonged Japanese ascendency, rather than with them still reeling from the physical and psychological shock of Midway. This will probably cause it to be delayed until a far more favourable balance of forces is obtained than OTL's invasion of Guadalcanal, and to be pushed less forcefully. This will exacerbate point 2).
The end result is that I can't see the counter-offensive beginning until the first half of 1944, when the Pacific Fleet has been built up to sufficient strength to be confident of defeating Kido Butai in any likely action. This puts back the timeline approximately eighteen months from OTL's Guadalcanal, and three months from Tarawa. I believe the former is the more realistic delay, given the vital role the battles in the Solomons played in degrading the strength and capability of the Japanese forces to resist the Central Pacific offensive. Adding in the additional advance required and the far stronger Japanese starting point (say, another six months) I believe that TTL's advance would reach the same culminating point as OTL's (i.e. preparing for the invasion of the Home Islands) by approximately June 1947. If Downfall proves necessary, that pushes the end of the war back to at least March-April 1948 (assuming TTL Olympic is 1 Nov 47 and Coronet 1 March 48).
There are three major factors that aren't or may not be affected by this POD.
The first is that it won't delay the Manhattan Project, which will still produce a deployable atomic weapon by July 1945. This means that it is possible that the first Superfort raids on Japan (approximately November 1946, assuming the rule above of OLT+2 years) will be nuclear, and will bring an end to the war as OTL. On the other hand, Truman (assuming this this chain of catastrophes doesn't kill Roosevelt early, cost him the '44 election, or change his thinking regarding his running-mate) may be advised or want to hold off launching nuclear raids until Japanese air defences have been degraded.
The second is August Storm. Despite the US still being half-way across the Pacific at this point, I believe it's still likely that the Russians will come stomping across the Manchurian border in August 1945, perhaps moreso, as they'll have a plenty of time to grab what they want before the US shows up. Whether this will by itself trigger a Japanese collapse probably largely depends on your view of what caused the OTL surrender, but in this case it will probably occur at either the same time or shortly after the invasion of the Marianas. What Stalin does next is anyone' guess - he may try and launch an invasion of the home islands or he may content himself with the Kuriles and turn his attention to the rest of the Japanese army in China.
The third wildcard is the submarine campaign. While I do not believe it could end the war on its own, it would effectively isolate the home islands, and accelerate the fall of bypassed outlying positions by reducing the defenders to a state of starvation. Whether it could do so to a sufficient degree to accelerate the progress of the main offensive is highly questionable.