The fact that you reached Dual 4-Back so quickly is quite evident of
your high IQ (and I don't mean that sarcastically -- most of us took a
LOT longer than 2 weeks to get to where we could routinely obtain
passing scores on dual 4-back).
Keep in mind that dual 4-back requires that either all four
immediately previous position elements or all four immediately
previous sound elements be remembered (and updated accordingly as new
trials are presented) to even stand a *chance* at having a better
score than that which could be obtained from the small probability
associated with correctly guessing an n-back stimuli match (seemingly)
without having any particular intuitively-derived inclination for the
correct answer.
Depending on your "chance of guaranteed match" setting (which I am
assuming are the default or more difficult), there is most likely
between a 12.5% (one in 8) and 32.5% (almost one in 3) chance that any
presented stimuli (eg. sound element or position element) will, "in
the long run" (with reference to mathematical probability), match the
stimuli which was presented n=4 positions ago. I believe the default
settings are (in the config.ini file which can be modified if you open
up the Brain Workshop package) chance_of_guaranteed_match = 0.20
which, in addition to the inherent 1/8 (=12.5%) chance due to there
being 8 positions and 8 letters/digits/etc (an 8-element subset is
chosen when a game starts), means you would have a 0.20 + 0.125 =
0.325 --> 32.5% chance of getting the correct answer if we assume it
is possible to have no "inclination" whatsoever to go on.
Long story, short, assuming you have the default settings, even if you
know the correct answer 100% of the time, only "on average" ("in the
long run, etc...") will you be rewarded for doing so with a correct
match. I am assuming you are using Manual Mode here, as Jaeggi mode
has a different and much easier scoring, which I don't think would be
causing you problems.
With a 32.5% chance of there being a position match and a 32.5% chance
of there being an auditory match, there is only a ~ 10.6% (0.325 *
0.325 = .106) chance of there being *both* position and auditory
matches, *however*, if you want to score high on dual n-back
(regardless of n-back level), you will have to remember the auditory
and position matches each time (eg. you can, for example, multiply
the .106 chance of a dual match by the percentage of the time (eg. 50%
--> .106 * .05 = .0053 = 5.3%) you believe you have been missing the
double match (be it one element or both elements) to obtain an
approximation for how much your score is getting hurt by these errors
(might need to increase/decrease that figure which is currently
assuming you correctly match on average one out of two stimuli on dual
matches). You could do the same for the auditory and position elements
and clearly are a smart guy, so I won't waste my time elaborating more
on this since I'm sure I already did far more than you would have most
likely needed (but did so since you're new to n-back and maybe haven't
"explored" the config.ini file yet and thus weren't sure how exactly
things were scored, so I hope this helped if that was the case).
Because the long run is irrelevant to any particular game, scores can
get hit pretty hard on occasion if you end up with a more difficult
session than the laws of probability would have "predicted" you to
likely obtain. One thing you can do if you have been noticing some
rather large game to game score discrepancies would be to use a larger
number of trials per session -- eg. make the laws of probability work
in your favor by making these subjectively or objectively more
difficult sessions rarer events. This might also apply if you're one
of the many members who are better at position or sound elements and
routinely have a noticeable scoring difference between the two -- a
lot of your stronger element being "correct matches" in a particular
game will make it easier for you [though perhaps more difficult for
someone else with different strengths].
I've been playing for a couple years now and spend most of my dual n-
back time playing Dual 4-Back with modified settings
(chance_of_guaranteed_match reduced to 0.05 and chance_of_interference
increased to 0.30 which make it considerably more difficult), however,
after all this time, I still can "only manage" around 75% on Dual 5-
Back with the default settings. Polar, a member of this board,
conducted a poll awhile back on members of this group and the "average
member" was playing either dual-4-back and dual-5-back. I can't
remember what the average number of months played was, but I believe
it was 9 months (someone correct me if I'm wrong here), so you're
doing progressing very nicely. Most members (who have shared their
opinion on this matter) have found successful performance at n-back
levels 4,5, and 6 to have required more time investment than other
lower or higher levels. Most people find N=2 to be pretty easy, people
may or may not find N=3 easy after a few days/weeks depending on their
working memory capacity, and most people will need to put in some real
time in order to succeed at levels 4+, since that capacity [whatever
it may be that is holding them back -- eg. working memory, "chunking
ability" (more relevant in modes incorporating 3/4+ stimuli),
executive functioning (eg. attention/focus), "g" (certainly no worries
about the problem being the latter in your case lol), etc] just wasn't
there initially.
What you may find useful to do is to count how many incorrect answers
you provide in a particular game -- dividing that by the number of
trials should give you a better understanding of how often you are
actually making mistakes, since 25 trials --> 25 possible sound + 25
possible position matches (since even if many trials aren't dual
matches, Brain Workshop will let you register every trial as such
anyway [resulting in incorrect answer(s)]. Out of all this, you'll
probably get around 10-20 (just a ballpark figure, due to the random
nature associated with the process) stimuli which will match the nth
trial preceding them. So technically, it's possible (though unlikely)
that a given game could have 16 n-back matches (let's assume 8 sound +
8 position) and you could get all of these wrong (because you didn't
register any of them as having occurred) but actually get everything
else right (eg. on the other 34 trials, you never mistakenly entered
any of these trials as having n-back stimuli matches). The game I just
described is actually quite real (it's what would happen if you did
nothing at all) and the score you obtain --zero percent-- will be no
different whether you started the game and went out for a coffee or
actually tried your best and had a very respectable performance but
incredibly bad luck (since making no mistakes on the other 34 trials
would mean you correctly identified the first 3 out of every 4 trials
as not being matches and would suggest you're probably capable of
near-100% performance on dual 3-back).
You may find increasing or decreasing the interference and guaranteed
match settings (respectively) to be helpful despite it seeming
somewhat counterintuitive that one should make the task even harder
than one is already finding it. With a lower guaranteed match setting,
the sound and position elements will be more random (eg. less
"unearned matches" due to a more random [and thus less predictable]
sequence). Similarly, higher interference settings will make make it
more difficult for our brain to try and compensate for any
difficulties we are having with the task at hand -- eg. if we treat
the dual n-back play area as a 3x3 matrix, we could view it as there
being 8 possible entries and associate position elements as having
occurred in row/column 1/2/3. If the last n=4 stimuli were all in row
1 and 2, you don't even need to remember where they were to know that
any position elements in row 3 will not be matches for the next 4
trials. Opinions are fairly divided on this board about whether
strategizing is helpful (eg. certainly for game scores) or harmful
(making the task "easier", though many will say that you could just
"bump up" the n-back level in such a case). I'll try not to reopen
that can of worms, however I try my best to avoid using more elaborate
strategies such as the one I just mentioned (some people use much more
"seemingly intuitive" strategies such as rehearsing the letters are
trying to track the positions with their eyes which I think is very
helpful when trying to play at or above your otherwise maximum n-back
level). The interference setting will generate "near-matches" (and is
explained better than I possibly can in the config.ini file in the
interference settings section) which should give you a pretty good
idea of whether you're genuinely flexing (and training) your cognitive
muscle, or merely getting better at playing an academic research-
backed brain game using methods which may or may not have been used by
participants in the published research.