#[doc = " https://github.com/hatoo/competitive-rust-snippets"]
#[allow(unused_imports)]
use std::cmp::{max, min, Ordering};
#[allow(unused_imports)]
use std::collections::{BTreeMap, BTreeSet, BinaryHeap, HashMap, HashSet, VecDeque};
#[allow(unused_imports)]
use std::io::{stdin, stdout, BufWriter, Write};
#[allow(unused_imports)]
use std::iter::FromIterator;
#[macro_export]
macro_rules ! chmax { ( $ x : expr , $ ( $ v : expr ) ,+ ) => { $ ( $ x = std :: cmp :: max ( $ x ,$ v ) ; ) + } ; }
#[macro_export]
macro_rules ! chmin { ( $ x : expr , $ ( $ v : expr ) ,+ ) => { $ ( $ x = std :: cmp :: min ( $ x ,$ v ) ; ) + } ; }
#[macro_export]
macro_rules ! max { ( $ x : expr ) => ( $ x ) ; ( $ x : expr , $ ( $ xs : expr ) ,+ ) => { std :: cmp :: max ( $ x , max ! ( $ ( $ xs ) ,+ ) ) } ; }
#[macro_export]
macro_rules ! min { ( $ x : expr ) => ( $ x ) ; ( $ x : expr , $ ( $ xs : expr ) ,+ ) => { std :: cmp :: min ( $ x , min ! ( $ ( $ xs ) ,+ ) ) } ; }
#[macro_export]
macro_rules ! dvec { ( $ t : expr ; $ len : expr ) => { vec ! [ $ t ; $ len ] } ; ( $ t : expr ; $ len : expr , $ ( $ rest : expr ) ,* ) => { vec ! [ dvec ! ( $ t ; $ ( $ rest ) ,* ) ; $ len ] } ; }
#[allow(unused_macros)]
macro_rules ! debug { ( $ ( $ a : expr ) ,* ) => { eprintln ! ( concat ! ( $ ( stringify ! ( $ a ) , " = {:?}, " ) ,* ) , $ ( $ a ) ,* ) ; } }
#[macro_export]
macro_rules ! input { ( source = $ s : expr , $ ( $ r : tt ) * ) => { let mut parser = Parser :: from_str ( $ s ) ; input_inner ! { parser , $ ( $ r ) * } } ; ( parser = $ parser : ident , $ ( $ r : tt ) * ) => { input_inner ! { $ parser , $ ( $ r ) * } } ; ( new_stdin_parser = $ parser : ident , $ ( $ r : tt ) * ) => { let stdin = std :: io :: stdin ( ) ; let reader = std :: io :: BufReader :: new ( stdin . lock ( ) ) ; let mut $ parser = Parser :: new ( reader ) ; input_inner ! { $ parser , $ ( $ r ) * } } ; ( $ ( $ r : tt ) * ) => { input ! { new_stdin_parser = parser , $ ( $ r ) * } } ; }
#[macro_export]
macro_rules ! input_inner { ( $ parser : ident ) => { } ; ( $ parser : ident , ) => { } ; ( $ parser : ident , $ var : ident : $ t : tt $ ( $ r : tt ) * ) => { let $ var = read_value ! ( $ parser , $ t ) ; input_inner ! { $ parser $ ( $ r ) * } } ; }
#[macro_export]
macro_rules ! read_value { ( $ parser : ident , ( $ ( $ t : tt ) ,* ) ) => { ( $ ( read_value ! ( $ parser , $ t ) ) ,* ) } ; ( $ parser : ident , [ $ t : tt ; $ len : expr ] ) => { ( 0 ..$ len ) . map ( | _ | read_value ! ( $ parser , $ t ) ) . collect ::< Vec < _ >> ( ) } ; ( $ parser : ident , chars ) => { read_value ! ( $ parser , String ) . chars ( ) . collect ::< Vec < char >> ( ) } ; ( $ parser : ident , usize1 ) => { read_value ! ( $ parser , usize ) - 1 } ; ( $ parser : ident , $ t : ty ) => { $ parser . next ::<$ t > ( ) . expect ( "Parse error" ) } ; }
use std::io;
use std::io::BufRead;
use std::str;
pub struct Parser<R> {
    reader: R,
    buf: Vec<u8>,
    pos: usize,
}
impl Parser<io::Empty> {
    pub fn from_str(s: &str) -> Parser<io::Empty> {
        Parser {
            reader: io::empty(),
            buf: s.as_bytes().to_vec(),
            pos: 0,
        }
    }
}
impl<R: BufRead> Parser<R> {
    pub fn new(reader: R) -> Parser<R> {
        Parser {
            reader: reader,
            buf: vec![],
            pos: 0,
        }
    }
    pub fn update_buf(&mut self) {
        self.buf.clear();
        self.pos = 0;
        loop {
            let (len, complete) = {
                let buf2 = self.reader.fill_buf().unwrap();
                self.buf.extend_from_slice(buf2);
                let len = buf2.len();
                if len == 0 {
                    break;
                }
                (len, buf2[len - 1] <= 0x20)
            };
            self.reader.consume(len);
            if complete {
                break;
            }
        }
    }
    pub fn next<T: str::FromStr>(&mut self) -> Result<T, T::Err> {
        loop {
            let mut begin = self.pos;
            while begin < self.buf.len() && (self.buf[begin] <= 0x20) {
                begin += 1;
            }
            let mut end = begin;
            while end < self.buf.len() && (self.buf[end] > 0x20) {
                end += 1;
            }
            if begin != self.buf.len() {
                self.pos = end;
                return str::from_utf8(&self.buf[begin..end]).unwrap().parse::<T>();
            } else {
                self.update_buf();
            }
        }
    }
}
#[allow(unused_macros)]
macro_rules ! debug { ( $ ( $ a : expr ) ,* ) => { eprintln ! ( concat ! ( $ ( stringify ! ( $ a ) , " = {:?}, " ) ,* ) , $ ( $ a ) ,* ) ; } }
#[doc = " https://github.com/hatoo/competitive-rust-snippets"]
const BIG_STACK_SIZE: bool = true;
#[allow(dead_code)]
fn main() {
    use std::thread;
    if BIG_STACK_SIZE {
        thread::Builder::new()
            .stack_size(32 * 1024 * 1024)
            .name("solve".into())
            .spawn(solve)
            .unwrap()
            .join()
            .unwrap();
    } else {
        solve();
    }
}
#[doc = "lower,upper are inclusive range"]
pub struct BinarySearch<F> {
    pub p: F,
    pub lower: i64,
    pub upper: i64,
}
impl<F: Fn(i64) -> bool> BinarySearch<F> {
    #[doc = "O(log(upper-lower))"]
    pub fn lower_bound(&self) -> i64 {
        let lower = self.lower;
        let upper = self.upper;
        assert!(lower <= upper);
        let mut lb = lower - 1;
        let mut ub = upper + 1;
        while ub - lb > 1 {
            let mid = (lb + ub) / 2;
            if (self.p)(mid) {
                ub = mid;
            } else {
                lb = mid;
            }
        }
        let latter = ub;
        latter
    }
}
fn solve() {
    input!{
        n:usize,
        a:[i64;n]
    }
    let mut a = a;
    a.sort();

    let mut cum = CumSum1::new(n);
    for i in 0..n {
        cum.set(i, a[i]);
    }
    cum.build();

    let mut maxv = 0;

    // 1つとるケース
    for i in 0..n {
        let m = a[i];
        let bs = BinarySearch {
            lower: 0,
            upper: min(i as i64, (n-i-1) as i64),
            p: |w: i64| {
                if w == 0 {
                    return false
                }
                let w = w as usize; 
                let l = i-w;
                let r = n-w;
                a[r]+a[l]<2*m
            } 
        };
        let r = bs.lower_bound() as usize; 
        let l = r - 1;
        let v = cum.query(i-l, i) + cum.query(n-l, n) - 2 * l as i64 * m;
        chmax!(maxv, v);
    }
    // 2つとるケース
    for i in 0..n-1 {
        let m0 = a[i];
        let m1 = a[i+1];
        let bs = BinarySearch {
            lower: 0,
            upper: min(i as i64, (n-i-2) as i64),
            p: |w: i64| {
                if w == 0 {
                    return false
                }
                let w = w as usize;
                let l = i-w;
                let r = n-w;
                a[r]+a[l] < m0+m1
            },
        };
        let r = bs.lower_bound() as usize;
        let l = r - 1;
        let v = cum.query(i-l, i) + cum.query(n-l, n) - l as i64 * (m0+m1);
        chmax!(maxv, v);
    }

    println!("{}", maxv);
}
// とりあえずN^2解法
fn solve_n2() {
    let out = stdout();
    let mut out = BufWriter::new(out.lock());
    input!{
        n:usize,
        a:[i64;n]
    }
    let mut a = a;
    a.sort();

    let mut cum = CumSum1::new(n);
    for i in 0..n {
        cum.set(i, a[i]);
    }
    cum.build();

    // kが決まってる時の最適値は計算出来る
    // O(N)
    let calc = |k: usize| {
        if k%2 == 1 {
            // 奇数ケース
            let mut maxv = 0;
            for i in 0..n {
                let m = a[i];
                let L = (k-1)/2;
                let R = (k-1)/2;
                
                if i+R >= n { continue; }
                if i<L { continue; } // i-L < 0

                let mut v = cum.query(i-L,i) + m + cum.query(n-R,n);
                v -= k as i64 * m;
                chmax!(maxv, v);
            }
            maxv
        } else {
            let mut maxv = 0;
            for i in 0..n-1 {
                let m0 = a[i];
                let m1 = a[i+1];
                let L = (k-2)/2;
                let R = (k-2)/2;
                if i+1+R >= n { continue; }
                if i<L { continue; }

                let mut v = cum.query(i-L, i) + m0 + m1 + cum.query(n-R, n);
                v -= (m0+m1) * k as i64 / 2;
                chmax!(maxv, v);
            }
            maxv
        }
    };

    let mut v = vec![];
    for k in 1..n+1 {
        v.push(calc(k));
    }

    let mut maxv = 0;
    for x in v {
        chmax!(maxv, x);
    }
    println!("{}",maxv);
}
struct CumSum1 {
    base: Vec<i64>,
    dp: Vec<i64>,
}
impl CumSum1 {
    fn new(n: usize) -> CumSum1 {
        CumSum1 {
            base: vec![0; n],
            dp: vec![],
        }
    }
    fn add(&mut self, i: usize, x: i64) {
        self.base[i] += x;
    }
    fn set(&mut self, i: usize, x: i64) {
        self.base[i] = x;
    }
    fn build(&mut self) {
        let n = self.base.len();
        let mut dp = vec![0; n + 1];
        let mut acc = 0;
        for i in 0..n {
            acc += self.base[i];
            dp[i + 1] = acc;
        }
        self.dp = dp;
    }
    #[doc = "[i,j)"]
    fn query(&self, i: usize, j: usize) -> i64 {
        self.dp[j] - self.dp[i]
    }
}